Why We Are Allocating Capital On Autopilot

In recent posts (here and here) I explored my view that today’s markets are systematically mispricing risk. My analysis isn’t exactly rocket science. So why does this mispricing persist? Why does everyone shrug their shoulders and say, “well, there is no alternative,” versus simply dialing back their exposures or hedging out some of the tail risk? At the very least, investors could increase the discount rates used in their valuations to correct for ultra-low riskfree interest rates and build in a greater margin of safety.

So why don’t they?

I would argue that more than anything, it is business and political pressures that drive this behavior. Importantly, I don’t believe this mispricing of risk is irrational. Rather, I believe decisions that seem rational on a micro level have led to irrational behavior in the aggregate. Investors are simply behaving how they are incentivized to behave–as a herd.

Here are my reasons:

Institutional investors must remain  invested. If you are a mutual fund manager or a hedge fund manager or venture capitalist, good luck explaining to your investors why you are sitting on a portfolio that is 40% cash. Many investors are loathe to stick with a manager who sits on a cash hoard for an extended period. Particularly in a buoyant market where cash will drag on returns. There is a sound rationale for this: the investor is perfectly capable of allocating to cash or hedging market risk on his own. Why pay some asset manager fees to sit on cash? While this makes plenty of sense from a business perspective, it makes no sense at all to an investing purist. The purist takes risk when the market is rewarding her for it and pares risk when the market is not rewarding her for it. Portfolios should be positioned more aggressively when markets are dislocated and prices are bombed out. They should be positioned more conservatively when valuations are high and expected returns are low.

Institutional investors are afraid to look different from their peers. Career risk drives a great deal of behavior in financial markets. It is the reason so many mutual funds look so similar to their benchmarks. This positioning makes no sense to a purist concerned with absolute returns. Yet it is perfectly rational for the mutual fund manager who will be fired if he drops into the fourth quartile of performance for a trailing 3-year period. Likewise for pension funds and endowments with trustees who may be penalized politically for contrarian positioning.

All investors have return hurdles to meet. If you are an individual or pension fund there is a certain rate of return that will allow you to fund your projected future liabilities. If you are an endowment or foundation there is some spending rule governing portfolio withdrawals, usually based on long-run capital market expectations. Altering these hurdles is a big deal. Reducing expected returns means pensions and individuals will have to save more to fund future liabilities. Endowments and foundations may have to cut financial support for certain programs. This can be psychologically devastating for individuals and extremely embarrassing for institutions. It is a powerful incentive for investors to take a “glass half full” view of the future, even if it is ultimately self-deluding and counterproductive.

Perhaps the most significant advantage you can get in the markets is what Ben Carlson calls organizational alpha. Put simply, this is the flexibility to do what others can’t, or won’t, as a result of business and political pressure. It is the freedom to switch off the autopilot and deviate from the pre-established flight plan.

More On S&P 500 PE Ratios

As follow up to my last post regarding cost of equity and valuation multiples for the S&P 500, here is a chart showing the implied steady state PE versus the actual PE for the index since 1961.

Data & Calculation Sources: Aswath Damodaran & Michael Mauboussin

The actual multiple typically plots above the steady state estimate. This is to be expected since the market is typically assigning some value to future earnings growth, and for simplicity’s sake my steady state multiple calculation does not factor in future growth.

By calculating the steady state multiple in this way you can easily visualize how the market is valuing future growth at a point in time. When the actual multiple is far above the steady state multiple, as in the late 1990s, the market is assigning a high value to future growth. Obviously the value of future growth can swing around violently depending on investor sentiment. In fact, exploiting this tendency for the market to overvalue and undervalue future growth is the lynchpin of Ben Graham style value investing. If you buy a stock at a low steady state valuation, yet have correctly discerned there are future growth opportunities not reflected in the market price, you get a free call option on the future.

The impact of ultra-loose monetary policy shows up very clearly in this chart in 2009. The steady state multiple shoots up dramatically in 2009 as interest rates (and thus discount rates) hit historic lows. What is interesting to me about this chart is how long the actual multiple remained below the steady state multiple, almost as if the market “realized” the discount rate had been artificially manipulated and refused to play ball. Again, that speaks to the power of investor sentiment.

Superficially, it appears as though the market valuation has finally “caught up” with its steady state value and has room left to run (remember, the steady state model isn’t pricing any growth). However, as discussed in the previous post, the steady state multiple has risen due to a low discount rate. So what I think investors need to think long and hard about today is whether we are systematically mispricing risk (spoiler alert: I think we are).

That said, I don’t think investors are mispricing risk because they are stupid, or even because they are greedy. In fact, I think they are acting rationally in the face of unappealing alternatives. But that is a subject to explore in future posts.

It’s The Risk You Don’t See That Kills You

I want to share a chart I believe is of paramount importance to asset allocation and valuation in today’s environment. It graphs the implied cost of equity for the S&P 500 alongside the steady state earnings multiple that number implies. The underlying data is from NYU Professor Aswath Damodaran and the steady state calculation is taken from a paper by Credit Suisse’s Michael Mauboussin.

Professor Damodaran estimates the implied cost of equity for the market by solving for the discount rate that sets the present value of projected S&P 500 dividends equal to the current level of the index. There are plenty of quibbles you can raise with this simple model (as with all models it is a dramatic oversimplification of messy reality). However, I don’t believe quibbles diminish the key insight.

Namely: the implied valuation of the equity market moves inversely to its implied cost of equity. This should make intuitive sense to anyone who has ever discounted a cash flow. If you were wondering, the calculated correlation is around -.90.

Here is my chart:

Data & Calculation Sources: Professor Aswath Damodaran & Michael Mauboussin

What this chart makes abundantly clear is that over the past thirty years the justified steady state multiple for the S&P 500 has crept steadily upward. Not so much because the risk premium has contracted but because the T-Bond rate (used as proxy for the riskfree rate of interest) has come down dramatically.

Now, that is a fairly superficial observation. Why does it matter?

Reason #1: A popular contrarian narrative in the markets is that central bankers have artificially suppressed interest rates, and that absent their interventions the “natural” rate of interest would be higher (implying a higher discount rate and thus lower sustainable valuation multiples). The key risk to this thesis is that low rates are not some exogenous happening imposed on the market by a bunch of cognac swilling technocrats, but rather a consequence of secular shifts in the global supply and demand for funds. Specifically, that these days there is a whole boatload of money out there that needs to be invested to fund future liabilities and too few attractive investment opportunities to absorb it all.

If low rates are actually a function of the supply and demand for funds, it doesn’t ultimately matter what central bankers intend do with monetary policy. Market forces will keep rates low and elevated valuations will remain justified.

Reason #2: Capital is allocated based on the opportunity set across asset classes. A 5% implied IRR on the S&P 500 may suck wind relative to “normalized” risk, but it is better than what you can expect in bonds. As long as equity looks like “the least bad alternative” valuations will remain elevated. If you are trying to short the market into this dynamic you are going to have a brutal go of it unless you are extremely fortunate in your timing. This is precisely why so many bearish investors have been wheeled out of the market on gurneys over the past 5 years.

Reason #3: By way of anecdotal evidence, sentiment is getting more and more bullish. Every day I am reading articles about the possibility of a market “melt-up.” If the market melts up it may narrow the implied risk premium and further reduce the implied discount rate. If this occurs, it leaves investors even more exposed to a double whammy: simultaneous spikes in both the riskfree interest rate and the risk premium. The years 1961 to 1980 on the chart give you an idea of how destructive a sustained increase in the discount rate can be to equity valuations.

My overall assessment of today’s markets is that perceived risk is low across the board. As a result investors are not generally being well-compensated for the real risks embedded in their portfolios.

Some Acknowledgements

This post is a synthesis of many arguments I have read over months and years. I therefore want to credit Josh Brown, John Hempton, Philosophical Economics, David Merkel and GMO for their tremendously insightful comments and analyses, which have helped me get to grips with the “real world” supply and demand fundamentals underlying today’s market valuations.

A Winner’s Curse

I have been having some interesting conversations recently regarding the latest trials and travails of cryptocurrency investors. The issue many of them are facing is what to do now having made returns of 5x, 10x, or more.

Do you let it ride and shoot for 1000x?

Do you lock in your profits now?

Something else?

In traditional markets, such as equity and fixed income, fundamental analysis helps with these issues. If you own Proctor & Gamble (PG) stock, and one day PG falls 50% for no reason other than that traders are bouncing the stock price around, you either: a) do nothing, or b) buy more. Although the market price has plummeted, there is no change in the intrinsic value of what you own (a slice of PG cash flows). In this case your valuation anchors you on what is important (intrinsic value) instead of the noise (the change in market price).

The challenge with cryptocurrency is that there is no intrinsic value for you to anchor on–at least not in the conventional sense. Holding forever and collecting your cash flows is not an option. There are no cash flows to collect. All you’ve got are supply and demand, and supply and demand are notoriously fickle over short time periods.

I have a pet theory that despite the meteoric rise in the price of Bitcoin, the average investor return is much, much lower. This would be consistent with investor behavior in traditional financial markets:


Of the municipal bond category, Morningstar’s Russ Kinnell wrote:

It’s surprising that the rather stable muni-bond fund group could be so misused, but it has been going on for a while. The problem here is that there are very risk-averse investors and a sector with scary headlines. The good news rarely makes headlines. Rather, investors hear about Puerto Rico’s crushing debt and Meredith Whitney’s ill-informed doomsday call. Those news events spurred muni investors to sell, which led to a drop in muni-bond prices and a spike in yields. Thus, they created a buying opportunity just as investors were fleeing. This speaks to the downside of trying to time the market and the benefit of staying focused on the long term.

Valuing A Bitcoin – Part III

Building off yesterday’s post today I will unveil a Bitcoin valuation.

Before we go any further I must emphasize that I am sharing this information as an intellectual exercise and for entertainment purposes only. This is not an investment recommendation and the output of this model should not be used to make investment decisions. You should consult with a financial advisor before making any investment decision. In the interest of full disclosure you should also know that I currently own neither cryptoassets nor exchange traded cryptoasset products (ETPs and ETNs).

The theoretical underpinning from this model is taken from Burniske and Tatar’s book, Cryptoassets. The authors propose adapting the Equation of Exchange (MV = PY) for valuing cryptocurrencies.

What the equation of exchange tells us is that the money supply times the velocity with which money circulates (left side) must equal the price level times real output (right side, a.k.a nominal output). So:

M = Money Supply

V = Velocity of Money

P = Price Level

Y = Real Output

I will apply the model to Bitcoin using data from blockchain.info. Many of my inputs will be rounded but I have always believed that perfect is the enemy of good when it comes to investing and valuation in particular. I am not sweating the small stuff. You are welcome to redo the work to two decimal points if spurious precision is your thing.

Anyway, we start with the supply of Bitcoin. This is easy. There will only ever be 21 million Bitcoins (unless of course the code is changed and that is a governance issue for the time being not a valuation issue). To be conservative I will assume all 21 million Bitcoin are in circulation for the valuation calculation.

The velocity of Bitcoin is a bit fuzzier but I can try to approximate the number using Bitcoin transaction data. According to the data Bitcoin transaction volumes are fairly stable oscillating around 200,000. We can annualize this by multiplying by 365 which equals about 73,000,000. We divide 73,000,000,000 by the current Bitcoin supply of about 16 million to get a velocity of about 4.56.

Price in USD is the variable we solve for. So we will pass over it for now.

With output we make a small adjustment and use output in USD terms as it will be easier to place our assumptions in context that way. This is about $1bn per day currently which we can annualize to about $365bn. That is estimated output today. What we need for our model is to also estimate the output at some point in the future. For the sake of this exercise let’s say in five years we think the USD equivalent transaction output for the Bitcoin network will be $1tn. This is a critical variable and some readers may think I am being overly conservative. Maybe so but do consider that this represents a compound annual growth rate of 112% a year.

We set up the model as follows:

21,000,000 x 4.56 = P ($1,000,000,000,000)

Solve for P using basic algebra and you get about .000096 BTC/USD. To make this number intelligible we take the reciprocal 1/.000096 to get USD/BTC which (using a spreadsheet for spurious precision) is about $10,443. That is a the estimated value of one Bitcoin five years from now.

For the final step we simply discount this price 5 years at our required rate of return. Since discount rate estimation is a pain and something of a guessing game in the best of times I like to simply choose a desired hurdle rate. For an asset like BTC I think 30% is reasonable given the risks and the immaturity of the asset class.

So discounting $10,443 for 5 years at 30% I estimate the value of one BTC today at $2,813. A summary of these calculations is included below.

Sources: Burniske & Tatar (model); Myself (calculations & tweaks)

Contrary to what some may think modeling is not about predicting the future. Rather it is about being explicit with your assumptions. This helps you test your assumptions for reasonableness. It also helps you identify the key variables you need to get right. Finally, it helps you build and maintain conviction in the face of market price volatility.

With Bitcoin here are the key variables:

  • How big can it get? -> How much “share” of global transaction volume will it take?
  • How long will it take to get there?
  • To what extent will it be used to transact versus as a store of value? The lower the velocity the more it is being used as a store of value and vice versa.
  • How much reward do you require given the risks?

You might disagree with my results and that is fine. However, I would ask you to consider where our views differ in the context of this model. Is it because you believe Bitcoin will get “bigger” and/or that it will get there “faster”? Is it because you think Bitcoin is less risky than I do?

I hope to update this valuation from time to time as Bitcoin evolves as an asset.

In closing, I would like to once again emphasize:

I am sharing this information as an intellectual exercise and for entertainment purposes only. This is not an investment recommendation and the output of this model should not be used to make investment decisions. You should consult with a financial advisor before making any investment decision. In the interest of full disclosure you should also know that I currently own neither cryptoassets nor exchange traded cryptoasset products (ETPs and ETNs).

Valuing Bitcoin – Part II

In my previous post on valuing Bitcoin I settled on supply/demand balance as the “least-bad” valuation model. I have been thinking more on how one might actual implement this in practice. The supply side is fairly straightforward. There are lots of free calculators that allow you to play with cost assumptions for Bitcoin miners. Now, there are probably going to be places in the world where an astute Bitcoin miner can arbitrage differences in electricity costs. But for now that’s splitting hairs.

The far trickier part is the demand side.

The reason is that while there are lots of use cases for Bitcoin, far and away the most prevalent is speculative trading. Therefore, if you take network activity at face value you are probably missing the fact that there is some reflexivity in those statistics. It’s basically a circular error problem. Speculative trading activity drives up network activity which drives up miner’s costs which causes the equilibrium price to rise. BUT, if speculative trading activity slackens (e.g. Bitcoin is in an asset bubble that deflates in the future) then the reverse will occur on the way down.

So in my view what you need to do is account for potential increases or decreases in speculative trading activity (and other kinds of activity) in your model. To do this you would need data that segments different transaction types.

The trick is finding that data.

As always this is not an investment recommendation. It is written for entertainment purposes only. As my thorough disclosure states very clearly, you should never make any investment decision based on something some random dude writes on the internet. Everything I am saying here could be wrong. In fact it is likely wrong. If you are looking for a recommendation on whether to own Bitcoin or any other cryptocurrency you should consult with a trusted financial advisor.

Bubble Logic

I have been reading a lot about cryptocurrencies lately. Before I go any further I must recommend two podcasts. They are:

Hash Power Part 1

Hash Power Part 2

These are impressive pieces of work. They were produced by Patrick O’Shaughnessy of the very excellent Investor’s Field Guide (you should also subscribe to Patrick’s regular podcast, Invest Like the Best). What I appreciated most about the podcasts is that they feature a number of intelligent, level-headed people trying their hardest to crack the toughest investment puzzle in recent history. These are not scumbag stock promoter types pumping and dumping ICOs. If you are interested in this kind of thing you should really check it out.

Anyway, back to the salient question(s). Are cryptocurrencies actually worth anything? If so, what are they worth?

I took a stab at this myself not too long ago. It was a useful exercise although it did not exactly end with concrete results. So despite having learned even more about blockchain and cryptocurrencies in the meantime, I remain stuck.

How am I supposed to invest in something that I cannot value?

Now, there is a pragmatic solution I have not really discussed (also mentioned by one of Patrick’s interviewees). That is, you can simply look at cryptocurrencies as call options (or, if you prefer less financial jargon, as lottery tickets). Viewed through the lens of portfolio construction this is far and away the best way of approaching the problem given the dramatic skew in the distribution of potential returns. Max downside is 100% of the original investment. And max upside is what? A 1000x gain? More? That is a pretty attractive option.

Yet it still doesn’t sit right with me. It feels too much like gambling. Which isn’t the worst thing in the world. I enjoy the occasional trip to the casino. However, conflating investing and gambling does not seem like a real answer. In fact it seems like bubble logic: gamble a little so you won’t miss out and regret it.

Valuing A BitCoin

This is not a post about what BitCoin is worth today, or will be worth in the future. No part of this should be construed as a recommendation to buy, sell, or hold BitCoins. If you landed here because you are wondering whether you should buy, sell or hold BitCoins, you need to do research elsewhere or consult with a trusted financial advisor, who can render an opinion based on your unique financial circumstances.

What this post is about are three different mental models one might use or adapt in analyzing the current price of BitCoin or another digital asset with similar characteristics.

First some background. I am fortunate to be friends with some very smart people with diverse sets of interests. We enjoy nerding out over similar topics: business strategy, financial analysis, technology, markets, entrepreneurship. If we can nerd out in person over a bottle of whiskey, all the better. Unfortunately now many years out of college several of us live in different cities. So we created a Slack group where we more or less maintain a running dialogue. Several of the longest running threads in our Slack group deal with cryptocurrencies.

I am not going to spend time or energy on background information in this post. There are smarter and more knowledgeable people than me all over the internet who can bring you up to speed on cyrptocurrencies. However, you should read the original BitCoin whitepaper. Primary sources matter. It doesn’t matter if you don’t understand every last detail. I certainly don’t.

The specific problem we were confronting on Slack was which type of mental model one might apply to a cryptocurrency like BitCoin to determine whether it is overpriced or underpriced at current market rates.

For example, there is an old saw about bank stocks that goes something like this: buy them at 1x book value and sell them at 2x book value. Whether 1x or 2x is the right multiple is irrelevant for the purposes of this discussion. The point is that the mental model for a bank or finance company revolves around the book value of its equity.

So what mental models might apply to cryptocurrency? We debated three on Slack:

1.       Purchasing Power Parity (PPP)

2.       Relative Value vs. Gold

3.       Supply/Demand Balance

Purchasing Power Parity

Purchasing power parity is the classic approach to assessing whether one currency is overvalued or undervalued relative to another (the key underlying assumption here is that BitCoin should be treated as a currency). PPP asserts that similar baskets of goods and services should trade for similar prices around the world. The Economist half-jokingly created The Big Mac Index in 1986 to analyze currency valuations around the world. Today The Economist has a dedicated web page devoted to Big Mac Index data, and the data has formed the basis for many books and academic studies. Here is what the Big Mac Index looks like today:

Source: The Economist

Pros: Intuitive, straightforward, much of the data is readily available. Captures the fact that BitCoin inflation should remain subdued versus fiat currencies over time (the supply of BitCoins is fixed at around 21 million whereas there is no limit on the amount of cash a government can print).

Cons: Goods and services are not priced “natively” in BitCoin. Starbucks does not say “a Grande Mocha costs $4 or 1 BTC” (indeed if that were the case BTC at $3,500 would seem wildly overvalued). This essentially blows up the PPP approach. Although personally I believe that if cryptocurrency truly goes mainstream, we will eventually get to the point where goods and services are priced natively in BTC and the PPP approach may have more analytical utility.

Relative Value vs. Gold

This was a simple thought experiment I conducted. Would I rather have an ounce of gold or 1 BitCoin? At the time I asked this question an ounce of gold was trading for about $1,300 and 1 BTC was trading for around $3,500. The intuition here is that gold and BitCoins essentially function as stores of value, where the market simply “agrees” on the value to assign each unit.

Pros: Exceptionally straightforward. Captures the notion that both gold and BitCoins have value more or less “because we agreed they have value” (if someone can explain to me why gold is considered to be a store of value, please drop me a line in the comments. Sorry gold bugs — I must confess I have never really understood why gold is so revered as a store of value).

Cons: Differences in unit measures. My comparison above makes 1 BTC look overvalued versus one ounce gold. But if you looked at the total value of the respective markets, BTC is worth an aggregate $50 billion while gold is worth an aggregate $9 trillion or so. On that basis, gold looks wildly expensive versus BTC.

We went around in circles on this for a long time. Personally, I still feel it is a useful thought experiment, though it raises hackles with others.

Supply/Demand Balance

A commodity like oil trades where supply and demand balance. On the demand side, human civilization requires a certain number of barrels of oil per day to function. This is something that can be estimated. On the supply side, it costs an exploration and production company a certain amount of money to get the oil out of the ground. So if it costs $50 to get a barrel worth of oil out of the ground, over the long term that sets a kind of floor for oil prices. Below that level, companies will not be able to cover their costs, more and more will go bankrupt and eventually the price will correct back toward an equilibrium level. This is of course a massive simplification but it illustrates the key principles.

Likewise, the BitCoin infrastructure is underpinned by the “miners” of the coins. Miners deploy the computing power that powers the network. They are incentivized to do this through the award of BTC for verifying transactions. The interesting wrinkle is that the system adjusts dynamically so that as more computing power is deployed on the network, it becomes more difficult to mine BTC (the awards get relatively smaller). At the same time, the miners’ costs rise, primarily due to ever-increasing electricity consumption. The reverse happens as computing power leaves the network (say, if unprofitable mining operations shut down). Several years ago you could mine BitCoin economically on a commercial laptop. Now profitable BitCoin mining requires significant scale and capital investment.

BitCoin’s supply side economics are relatively straightforward to understand at a high level. The demand side is more challenging. For me the most significant impediment to modeling the demand side is that the majority of BitCoin transaction activity currently appears to be speculative trading. If the data in the linked article is to be believed, less than 1% of transactions are related to actual payment processing, with exponentially higher volumes driven by trading activity. This creates some reflexivity as relates to “network fundamentals.” More trading activity -> more transactions -> implies higher demand. However, that self-reinforcing momentum can easily unwind again on the way back down. I do not have an easy answer for how to deal with this, but I have the sense that it is of critical importance.

Conclusions (Such That They Are)

If you are a BitCoin trading enthusiast, and you are still reading this, I imagine that you are thinking something along the lines of “this idiot spent 1,000 words on that!? I am no closer to understanding whether now is a good time to buy!”

To reiterate: this is not a post about what BitCoin is worth today, or will be worth in the future. No part of this should be construed as a recommendation to buy, sell, or hold BitCoins. If you landed here because you are wondering whether you should buy, sell or hold BitCoins, you need to do research elsewhere and consult with a trusted financial advisor, who can render an opinion based on your unique financial circumstances.

All useful analysis is rooted in finding the right questions to ask. Looking at a simple bank stock, for example, the right questions are along these lines: what is a fair estimate of book value per share? Is the quality of the loan book truly reflected on the balance sheet? If not, how do I adjust those numbers to more accurately reflect reality? Usually these questions will lead you to further questions centered on the breakdown of the bank’s loan book and the quality of its underwriting standards.

Similarly, the first step toward assessing whether BitCoin is overpriced or underpriced involves identifying the right questions to ask about its fundamentals.

What do you think?