Fiscal Inflation Targeting and the Cost of Large Government Debt Accumulation

cheney“You know, Paul, Reagan proved that deficits don’t matter.  We won the mid-term elections, this is our due.”

— Vice President Dick Cheney defending a second round of tax cuts against the objection of Treasury Secretary Paul O’Neill, shortly after the 2002 mid-term elections.

Over the last two decades, the Japanese economy has failed to generate healthy levels of inflation.  Some might wonder why that’s a problem.  It’s a problem because healthy inflation is the only reliable indication that an economy is fully utilizing the labor and capital resources available to it.  The fact that Japan has had no inflation, or has even deflated, confirms that it has been operating below its potential, at the expense of the living standards of its population.

What should Japanese policymakers do to restore inflation to healthy levels?  This question is extremely important, not only for the economic future of Japan, but for the economic future of the entire world.  The causes of persistently weak inflation in Japan aren’t entirely understood, but it’s possible–and likely–that they are tied to the effects of slowing population growth, aging demographics, and a growing scarcity of worthwhile investment opportunities, ways to add real value to the economy by creating new capacities that consumers will genuinely benefit from and be eager to spend their incomes on.  If that’s the case, then the entire world will eventually face the problems that Japan currently faces, as the entire world is on Japan’s demographic and developmental path.

For the past 2 years, Japanese policymakers have attempted to use unconventional monetary policy to stimulate inflation.  The Bank of Japan (BOJ) has set an explicit 2% target for the inflation rate, and has effectively promised to purchase whatever quantity of assets it needs to in order to bring the inflation rate to that target.  Unfortunately, asset purchases don’t affect inflation, and so the BOJ has essentially been wasting everyone’s time. In terms of the actual data, the policy has been a clear failure, with core Japanese YOY CPI running at a pathetic 0%, despite a two-year doubling of Japan’s already enormous monetary base.

The solution for Japan, and for any economy that is underutilizing its resources, is to implement an inflation targeting policy that is fiscal rather than monetary–what we might call “Fiscal Inflation Targeting.”  In Fiscal Inflation Targeting, the legislature sets an inflation target–e.g., 2% annualized–and gives the central bank the power to change the rates on a broad-based tax–for example, the lower brackets of the income tax–as needed to bring inflation to that target.  To address scenarios in which inflation chronically undershoots the target, the central bank is given the power to cut tax rates through zero, to negative values, initiating the equivalent of transfer payments from the government to the private sector.  Crucially, any tax cuts or transfer payments that the central bank implements under the policy are left as deficits, and the ensuing debt is allowed to grow indefinitely, with the central bank only worrying about it to the extent that it impedes the ability to maintain inflation on target (via a mechanism that will be carefully explained later).  Note that other macroeconomic targets, such as nominal GDP (nGDP), can just as easily be used in lieu of inflation.

A policy of Fiscal Inflation Targeting would be guaranteed to achieve its stimulatory goals. In terms of direct effects, tax cuts and transfers directly increase nominal incomes, and there is no level of inflation that cannot be achieved if nominal incomes are broadly increased by a sufficient amount.  In terms of indirect effects, having such a powerful tool to use would dramatically increase the economic impact of the central bank’s communications.  On the current approach, to stimulate at the zero lower bound, the central bank is limited to the use of balance sheet expansion, which works primarily by placebo, if it works at all.  But in Fiscal Inflation Targeting, the central bank would have the equivalent of real drugs to use. Economic participants would therefore have every reason to trust in its power, and to act as if its targets would be consistently met.

The insight that fiscal policy should be used to manage inflation, in the way that monetary policy is currently used, is not new, but was introduced many decades ago by the founders of functional finance, who were the first to realize that inflation, and not the budget, is what constrains the spending of a sovereign government.  Advocates of modern monetary theory (MMT), the modern offshoot of functional finance, notably Scott Fulwiller of Wartburg College, have offered policy ideas for how to implement a fiscally-oriented approach.

My view, which I elaborated on in a 2013 piece, is that the successful implementation of any such approach will need to involve the transfer of control over a portion of fiscal policy from the legislature and the treasury to the central bank.  Otherwise, the implementation will become mired in politics, which will prevent the government’s fiscal stance from appropriately responding to changing macroeconomic conditions.

There are concerns that such a policy would be unconstitutional in the United States, since only the legislature has the constitutional authority to levy taxes.  But there is no reason why the legislature could not delegate some of that authority to the Federal Reserve in law, in the same way that it delegates its constitutional authority to create money.  In the cleanest possible version of the proposal, the legislature would pass a law that creates a special broad-based tax, and that identifies a range of acceptable values for it, to include negative values–say, +10% to -10% of earned income below some cutoff.  The law would then instruct the Federal Reserve to choose the rate in that range that will best keep inflation on target, given what is happening elsewhere in the economy and elsewhere in the policy arena.

Ultimately, the chief obstacle to the acceptance and implementation of fiscal inflation targeting is the fear that it would lead to the accumulation of large amounts of government debt.  And it would, particularly in economies that face structural weakness in aggregate demand and that require recurrent injections of fiscal stimulus to operate at their potentials.  But for those economies, having large government debt wouldn’t be a bad thing.  To the contrary, it would be a good thing, a condition that would help offset the weakness.

The costs of large government debt accumulation are not well understood–by lay people or by economists.  In this piece, I’m going to try to rigorously work out those costs, with a specific emphasis on how they play out.  It turns out that there is currently substantial room, in essentially all developed economies that have sovereign control over credible currencies, to use expansive fiscal policy to combat structural declines in inflation, without significant costs coming into play.

The reader is forewarned that this piece is long.  It has to be, in order to make the mechanisms fully clear.  For those that want a quick version, here’s a bulleted summary of the key points:

  • Government debt accumulation increases the net financial wealth of the private sector. (Note: for convenience, from here forward, we will omit the “net” term, and use the terms “net financial wealth” and “financial wealth” to mean the same thing).
  • Increases in financial wealth can lead to increases in spending which can lead to increases in inflation.  The effects need not be immediate, but may only show up after conditions in the economy have changed such that the economy’s wealth velocity–the speed at which its total stock of financial wealth “circulates” in the form of expenditures–and its wealth capacity–its ability to store financial wealth without overheating–have risen and fallen respectively.
  • From a policy perspective, the way to reduce an economy’s wealth velocity and increase its wealth capacity is to raise its interest rate.  But when government debt is overwhelmingly large, interest rate increases tend to be either destabilizing or inflationary, depending on how the government funds the increased interest expense that it ends up incurring.
  • The countries that are the best candidates for fiscal inflation targeting are those that have structurally high wealth capacities–those that are able to hold large amounts of financial wealth without overheating, and that are likely to retain that ability indefinitely into the future.  Examples of such countries include the United States, Japan, the U.K., and the creditor countries of the Eurozone.

The piece is divided into two parts.  I begin the first part by specifying what counts as a “cost” in an economic policy context.  I then examine four myths about the costs of large government deficits: (1) That there can be no free lunch, (2) That large government deficits are unsustainable, (3) That large government deficits cause interest rates to rise, and (4) That large government deficits are necessarily inflationary.

In the second part, I explain the mechanism through which fiscal inflation targeting, and any policy approach that generates exceedingly large quantities of government debt, can sow the seeds of a future inflation problem.  I begin by introducing the concept of “wealth velocity”, which is a modification of the more economically familiar term “money velocity”, and “wealth capacity”, which is a concept analogous to “heat capacity” in thermodynamics. I then explain how changes in wealth velocity and wealth capacity over time can cause a stock of accumulated government debt that wasn’t inflationary to become inflationary, and how the the ability of monetary policy to appropriately respond can be curtailed by its presence.  I conclude the piece with a cost-benefit analysis of fiscal inflation targeting, identifying the countries in the world that are currently the best candidates for it.

Functional Finance and “Costs”: A Focus on the Well-Being of Persons

costs

Suppose that to keep the economy on a 2% inflation target, a government would have to run a large deficit–say, 10% of GDP–in perpetuity.  What would the cost of perpetually running such a deficit be?  Answer: the eventual accumulation of a large quantity of government debt.  But this answer begs the question.  What is the cost of accumulating a large quantity of government debt? Why should such an accumulation be viewed as a bad thing? What specific harm would it bring?

To speak accurately about the “cost” of large government debt, we need to be clear about what kinds of things count as real costs.  To that end, we borrow from the wisdom of the great economist Abba Lerner:

“The central idea is that government fiscal policy, its spending and taxing, its borrowing and repayment of loans, its issue of new money and its withdrawal of money, shall all be undertaken with an eye only to the results of these actions on the economy and not to any established traditional doctrine about what is sound or unsound.  This principle of judging only by effects has been applied in many other fields of human activity, where it is known as the method of science as opposed to scholasticism.  The principle of judging fiscal measures by the way they work or function in the economy we may call Functional Finance.” — Abba Lerner, “Functional Finance and the Federal Debt”, 1943.

In the context of economic policy, a real cost is a cost that entails adverse effects on the well-being–the balance of happiness and suffering–of real people, now or in the future.  A good example of a real cost would be poverty brought on by unemployment.  Poverty brought on by unemployment entails concrete suffering for the afflicted individuals.

Another good example of a real cost would be hyperinflation. Hyperinflation represents a nuisance to daily life; it undermines the enjoyment derived from consumption, forcing consumers to consume because they have to in order to avoid losses, rather than because they want to; it creates an environment in which financially unsophisticated savers end up losing what they’ve worked hard to earn; it makes contractual agreements difficult to clearly arrange, and therefore prevents economic parties from engaging in mutually beneficial transactions; it retards economic growth by encouraging allocation of labor and capital to useless activities designed to protect against it–e.g., precious metal mining. Each of these effects can be tied to the well-being of real people, and therefore each is evidence of a real cost.

In contrast, “not having our fiscal house in order” or “owing large amounts of money to China” or “passing on enormous debts to our children” are not real costs, at least not without further analysis.  They don’t, in themselves, entail adverse effects on the well-being of real people, now or in the future.  Their rhetorical force comes not from any legitimate harms they cite, but from their effectiveness in channeling the implied “moral guilt” of debt accumulation–its connection to short-termism, hedonism, selfishness, impulsiveness, recklessness, irresponsibility, and so on.  In that sense, they are like the prevailing rhetorical criticisms of homosexuality.  “But that’s gross!” is not a valid objection to consensual love-making activities that bring happiness to the participants, and that cause no harm to anyone else.  Similarly, “But we’re spending money we don’t have” is not a valid objection to fiscally expansive policies that improve the general economic well-being without attaching adverse short or long-term economic consequences.

None of what is being said here is meant to deny, off the bat, that large government debt accumulation can bring adverse economic consequences.  It certainly can.  But to be real, and to matter, those consequences need to involve real human interests–they can’t simply be empty worries about how a government’s finances are “supposed to be” run.

Myth #1 — There Can Be No Free Lunch

Imagine a primitive, specialized economy that trades not in money, but in promises. You are skilled at making food, and have extra leftovers from your recent meal to share; I am skilled at construction, and can build shelter–a hut–for you.  Your present hut is currently adequate, but it will eventually fall apart and need to be rebuilt.  So we make a deal. I will rebuild your hut, on your request, and, in exchange, you will give me your leftover food.

now later

Importantly, there’s a time lag between the delivery of my end of the deal and the delivery of your end.  You have food right there, ready to give, and I am ready to take it, now.  But the rebuilding of your hut is only going to take place later, when the need arises.  It follows that in this trade, an actual good–food, right here in front of both of us–is being exchanged for a potential good, a promise to do something at some point in the future, on request.

Now, let’s suppose that my appetite for food is enormous, and that I enter into similar deals with other people, in order to get food from them.  As I go on accumulating more and more hut-building debt, I think to myself, “This is great. I can eat as much as I want, and I don’t have to actually do any work.” Using this tactic, will I be able to secure a “free lunch” for myself–literally, a lunch that I will never have to repay?

The obvious answer is no.  The people that are accepting my promises as compensation aren’t idiots.  They are accepting them in order to one day use them.  If I issue more promises than I can reasonably make due on, then when the owners of those promises eventually come to me wanting to use them, wanting me to rebuild their huts, I’m not going to be able to deliver what I owe.  I won’t have sufficient time or resources to rebuild the shelters of all of the people that I’ve made promises to.  That’s where the perceived “free lunch” will fall apart.  For some, my lunches won’t have been free lunches–they will have been stolen lunches, lunches that I wrongly took without the prospect of repayment, and that I am sure to be retaliated against for having taken.

Our natural inclination is to extend this intuition to the operation of specialized economies that trade in money.  Suppose that there is a disabled homeless man that cannot find gainful employment, a way to contribute.  It’s not his fault–his circumstances are such that there just isn’t anything useful that he can do for anyone, no value that he can add to anyone’s life that would make anyone want to pay him anything.  That said, he has certain needs–food, shelter, clothing, medical attention, and so on.  From a general humanitarian perspective, we want those needs to be met.

As a society, how might we ensure that his needs are met?  The “fiscally honest” way would be to require anyone that earns income to give a portion of that income to the government through taxes, and to then have the government disburse the proceeds to him, and to others like him, to spend on basic necessities.  In imposing this requirement, we would be confiscating a portion of their potential consumption, which they’ve earned through their productive contributions, and which they own in the form of the money they are holding, and transferring that consumption to him.

But is that the only way to ensure that his needs are met?  What if instead of taking money from income earners, and giving it to him and to others like him, we were to simply create new money–print it up from nothing?  In printing new money and giving it to him, we would be giving him the ability to purchase the things he needs, without having to take away anyone else’s money–the undesirable part of the process that we would surely avoid if we could.

Our prior intuition, that there are no free lunches, enters the picture here, and causes us to search for a hidden cost in the approach–a consequence that we haven’t properly acknowledged or accounted for. Surely, things can’t be that easy, that we would be able to use money creation to entirely circumvent the need to actually part with the things that we give away to others.  In terms of what that cost actually is, we normally assume it to be inflation.  In creating new money and spending it, we reduce the value of existing money. In this way, we take wealth from those that are holding it in the form of money, and transfer it to those that aren’t.

But there’s a mistake in this thinking.  The seeds of the mistake were identified long ago, by two philosophers who are far more famous for other ideas: the British philosopher David Hume and the German philosopher Arthur Schopenhauer.

Beginning with Hume,

“It is also evident, that the prices do not so much depend on the absolute quantity of commodities and that of money, which are in a nation, as on that of the commodities, which come or may come to market, and of the money which circulates.  If the coin be locked up in chests, it is the same thing with regard to prices, as if it were annihilated; if the commodities be hoarded in magazines and granaries, a like effect follows.  As the money and commodities in these cases never meet, they cannot affect each other.”  — David Hume, Essays, Moral, Political, Literary, “Of Money“, 1741

Hume’s point was that a simple increase in the stock of money is not enough to cause an increase in prices.  To be inflationary, increases in the money stock have to lead to increases in spending.  If they do not lead to increases in spending, then they will not affect the balance of supply and demand–the balance that determines the trajectory of prices.

Schopenhauer summed up the second seed of the mistake in the following quotes:

“People are often reproached because their desires are directed mainly to money, and they are fonder of it than of anything else. Yet it is natural and even inevitable for them to love that which, as an untiring Proteus, is ready at any moment to convert itself into the particular object of our fickle desires and manifold needs. Thus every other blessing can satisfy only one desire and one need; for instance, food is good only to the hungry, wine only for the healthy, medicine for the sick, a fur coat for winter, women for youth, and so on. Consequently, all these are only relatively good. Money alone is the absolutely good thing because it meets not merely one need considered concretely, but all needs considered abstractly.” — Arthur Schopenhauer, The World as Will and Representation, Volume I, 1818.

“Money is human happiness in the abstract: he, then, who is no longer capable of enjoying human happiness in the concrete devotes his heart entirely to money.” — Arhtur Schopenhauer, Counsels and Maxims, 1851.

The mistake is to assume that anyone who exchanges the output of her time, labor, and capital for money does so because she eventually wants to use that money to consume the output of the time, labor, and capital of someone else.  If money had no value outside of its use in consumption, then this assumption might make sense.  It would be irrational for an individual to work for money that she wasn’t ever going to spend–she would essentially be working for free.  But in an economic system where money is the primary mode of trade, it acquires intangible value–value unrelated to the actual purchase of any concrete good or service.  It comes to represent abstract, psychological goods: happiness, accomplishment, success, optionality, ability, power, safety, security, status, respect, and so on.  A person may have accumulated enough of it to meet her actual future consumption needs many times over, but she will still seek more of it, in pursuit of those goods.

Suppose that Warren Buffet were to make a series of bad investments that were to cause the monetary market value of his wealth–currently, $72.3B–to permanently shrink by 99.9%, leaving “only” $72.3MM to his name.  The loss would not in any way affect his consumption or the consumption of his heirs, for neither he, nor they, are likely to put either amount to consumptive use.  But still, he would be extremely upset by it, and would go to great lengths to avoid it.  Why?  Because money holds intangible value to him, value that is unrelated to its actual use in funding his current and future consumption expenditures.

Behaviorally, he has been trained, from his youth, his days as a much poorer man, to respect money, and to never let it be wasted or left on the table. It is something that is supposed to be cared for, nurtured, grown over time–not shrunk.  To lose so much of it would therefore be frustrating and unpleasant for him, even if the loss made no difference to his lifestyle.

Why does Buffet care about money?  What value does it bring him?

  • Success, Achievement, Scorekeeping.  Money is the way that success and achievement are measured, kept score of–in Buffet’s case, success and achievement at the game of investing, a game that he enjoys playing, and that he wants to win at.  Money is the reward that makes investing a serious game, a game of real significance, that brings pleasure when played well.
  • Power, Freedom.  The knowledge that he can do and have any possible thing that he wants, whenever he wants it, in whatever quantity he wants it–up to infinity. That knowledge brings satisfaction, even if the underlying capacity will never be put to use.
  • Safety, Security.  The knowledge that his wants and needs will never go unmet, that his standard of living, and those of the people he cares about, will never fall to unwanted levels, that the causes he believes in will always have an able advocate in him.  That knowledge makes it easier for him to enjoy the things that he actually does partake in, removing the possibility–and therefore the fear–that his ability to partake in them might one day be compromised.
  • Social Status.  The respect of other people, who admire it as an amazing accomplishment, who rightly interpret it as a sign of his acumen and his value to society, and–let’s be honest–who want to be close to him because of that.  If its value were to fall dramatically, he would lose some of that admiration, that respect, that special attention that he gets from the world.  As a human being with pride, he would surely suffer at the loss.

Many successful individuals in our society that have accumulated large amounts of wealth, Buffet included, have pledged to give it all to charity.  But notice that the pledges are always pledges to give it all to charity at death, never to give it all to charity right now.  But why not give it all to charity right now, keeping only what is necessary to fund future consumption?  Because to do so would require parting with the intangible goods that it confers right now, and that it will continue to confer up until death.

From a Humean perspective, then, Warren Buffet’s $72.3B might as well be “locked up in chests”–it does not circulate in expenditures, and therefore does not affect prices.  That is precisely where the potential “free lunch” of a government deficit lies.  If money is printed anew and given to Warren Buffet, in exchange for work that he does for others, the money will go into his piggy bank, where it will have no effect on anything.  The work done will therefore have been provided at no cost to anyone–no cost to Buffet, no cost to the recipients, and no cost to the collective.  It will have been paid for by the intangible goods that attach to money, goods that the government can create for free.

Now, Warren Buffett is an extreme example.  But any person who puts income into savings, never to be consumed, is operating on the same principle.  And a large number of the participants in our global economy–who collectively control enormous amounts of wealth–do just that.  They put their income into savings, which they never end up consuming.  The fact that they do this is the reason that a significant portion of the world cannot find gainful employment, a way to contribute.  It’s also the reason that a free lunch is on the table, a free lunch that could solve the unemployment problem, if only policymakers knew that it was there to be taken.

Let’s make the point more precise.  There are two things that a person can do with income. First, spend it–use it to consume the economy’s output.  Second, save it. There are two ways that a person can save income.  First, by holding it as money (or trading it for an existing asset, in which case the seller of the asset ends up holding it as money). Second, by investing it–and by “investing it” I mean investing it in the real economy, using it to fund the construction of a new economic asset–a home, a factory, etc.–that didn’t previously exist.

Inflation occurs in environments where excessive demand is placed on the economy’s labor and capital resources–demand for which there is insufficient supply.  From the perspective of inflation, the only income that directly matters is income that is spent or invested–income that is used to put demand on the economy’s labor and capital resources. Income that is held as savings does not put demand on those resources, and therefore has no effect on prices.  It follows that if a government creates new money to finance a deficit, delivering that money as income to someone in the form of a tax cut, a transfer, or a direct payment in exchange for goods and services provided, and if the receiver of the income is not inclined to spend or invest it, but instead chooses to hold it idly in savings, in pursuit of the intangible goods that savings confer, then a free lunch is possible.  Everyone can benefit, without anyone having to sacrifice.

Now, free lunches aren’t on the table everywhere.  But in the economies of the developed world, where there are large output gaps and large overages in the demand for savings relative to the demand for investment, free lunches are on the table.  Unfortunately, many policymakers don’t understand how they work, and therefore haven’t been able to take advantage of them.

Myth #2 — Large Government Deficits Are Unsustainable

If a government runs a large deficit in perpetuity, the debt will grow to an infinitely large value, a value that the government won’t realistically be able to pay back.  But there’s nothing wrong with that.  Government debt is supposed to grow to infinity, along with all other nominal macroeconomic aggregates.  It isn’t supposed to ever be paid back.

In essentially every country and economy that has ever existed, nominal government debts have grown indefinitely.  They have never been fully paid back, only refinanced or defaulted on. The following chart shows the gross nominal debt of the U.S. Federal Government from 1939 to present (FRED: Gross Federal Debt, log scale).  As you can see, there was never a sustained period in which any substantial portion of it was paid down:

gfd

What matters is not the nominal quantity of debt that a government owes, but the ratio of that quantity to the economy’s nominal income, which is the income stream from which the government draws the tax revenues that it uses to service the debt.  That income stream also grows to infinity, therefore its ratio to government debt can stabilize at a constant value, even as the government continues to run deficits year after year after year.

Let d refer to the size of the primary government deficit (before interest expense is paid) as a percentage of GDP, let i refer to the average annual interest rate paid on the outstanding government debt, and let g refer to the economy’s nominal annual GDP growth, given as a percentage of the prior year’s GDP.  We can write a differential equation for the rate of change of the debt-to-GDP ratio over time.  Setting that rate equal to zero and solving, we get the following equation for the debt-to-GDP ratio at equilibrium:

(1) Debt / GDP = d / (g – i)

When we say that a deficit is sustainable, what we mean is that running it in perpetuity will produce a Debt-to-GDP ratio that stabilizes at some value, rather than a debt-to-GDP ratio that grows to infinity.  Per equation (1), any government deficit, run in perpetuity, will be sustainable, provided that the economy’s equilibrium nominal growth rate g exceeds the equilibrium interest rate i paid on the debt.

To illustrate the application of equation (1), suppose that to keep the economy on a 2% inflation target, with 4% nominal growth, a government would have to run a primary deficit of 5% of GDP in perpetuity.  Suppose further that the nominal interest rate that the central bank would have to set in order to control inflation under this regime would be 1%. At what value would the Debt-to-GDP ratio stabilize?

To answer the question, we note that d is 5%, g is 4%, and i is 1%.  Plugging these values into the equation, we get,

(2) Debt / GDP = 5% / (4% – 1%) = 5% / 3% = 166%

Now, let’s increase the interest rate i from 1% to 3%.  Plugging 3% into the equation, we get,

(3) Debt / GDP = 5% / (4% – 3%) = 5%/1% = 500%

As you can see, the Debt-to-GDP ratio is extremely sensitive to changes in the interest rate paid on the debt, particularly as that rate gets closer to the economy’s nominal growth rate.

The following chart shows the amount of time that it would take to get to the 500% Debt-to-GDP equilibrium, assuming a 200% Debt-to-GDP starting point (roughly, Japan’s current ratio, in gross terms).

dngdp

As you can see, it would take a very long time–more than 400 years.  Just to get to 300% Debt-to-GDP would take over 50 years.  It’s misguided, then, for policymakers to worry about the sustainability of deficits that will only need to be run, in worst case scenarios, for a decade or two.

Now, in countries such as Greece and Ecuador, where the government debt is denominated in an external currency, large accumulation of government debt can be dangerous.  The interest rate paid on the debt is set directly by lenders in the market, and therefore the equivalent of a fiscal “bank run” can develop, in which lenders, concerned with rising debt, demand higher interest rates in order to lend, which causes deterioration in the fiscal budget, which increases concern among lenders, which leads them to demand even higher interest rates in order to lend, which causes further deterioration, and so on.

Fortunately, the countries outside of the Eurozone that currently need fiscal stimulus–the U.S., Japan, and the U.K.–have sovereign control over the currencies that their debts are denominated in.  They can therefore set the interest rates on their debts as low as they want to, short-circuiting any attempted run on their finances.  Of course, the consequence of setting interest rates too low might be an inflation–a “run” in a different form–but a direct run, occurring in the form of lenders refusing to lend, can always be stopped.  This point will be addressed in further detail in subsequent sections.

Myth #3 — Large Government Deficits Lead to Rising Interest Rates

It’s often assumed that if a government accumulates debts in large amounts, that the interest rate that the market will demand in order to lend to that government will rise, not only in nominal terms, but in real terms. This assumption is based on two feared dynamics:

  • Positive Feedback Loop: Rising debt implies rising default risk, which causes lenders to demand higher interest rates in order to lend, which worsens the fiscal picture, which increases the default risk, which causes lenders to demand even higher interest rates in order to lend, and so on.
  • Excessive Supply: The government has to find people willing to hold its debt in asset portfolios.  As its debt increases, finding the needed quantity of willing holders becomes more difficult.  Higher interest rates then have to be offered in order to attract such holders.

These concerns are refuted by actual experience.  With the exception of Greece, a country tied down to an external monetary standard from which it is expected to eventually exit, the countries with the highest debt levels in the world relative to GDP–the U.S., Japan, the U.K., and the rest of the Eurozone–are able to borrow at the lowest interest rates in the world.  This relationship isn’t a recent phenomenon–it was observed in past eras as well.  The U.S. and the U.K. accumulated very large government debts relative to GDP in the 1940s. But interest rates during that decade stayed very low–indeed, at record lows–and remained low for more than a decade afterwards.

High debt levels relative to GDP and low interest rates are often observed together because they share a number of the same potential causes.  Persistently weak nominal economic growth, for example, calls for policymakers to lower interest rates.  It also demands more fiscal stimulus, fueling a faster increase in government debt.  The debt-to-GDP ratio itself grows faster because the growth in the denominator of the expression stalls, as debt continues to be added on to the numerator.

Those who fear that large government debt accumulation will put upward pressure on interest rates do not fully understand how interest rates work.  Interest rates in a given currency are ultimately determined by the issuer of that currency.  The mechanisms can be different in different types of systems, but the issuer usually controls the interest rate by expanding or reducing the quantity of loanable funds in the system, which moves the interest rate down or up in accordance with the dynamic of supply and demand.  If the issuer wants to, it can go so far as to create and lend out money directly, at whatever rate it wants, to whomever it wants.

Rising interest rates, then, are not a meaningful risk for a government that owes debt in a currency that it issues.  The government itself gets to determine what its interest rate is going to be.  In practice, the government is going to set interest rates at the minimum level that keeps inflation on target.  Fortunately, in an economy that is suffering from structural weakness in aggregate demand, that level will tend to be very low, allowing for the servicing of a large debt.

The best way to illustrate these points is with a concrete example.  Suppose that in watching the national debt expand indefinitely, lenders in the U.S. were to become afraid of an eventual government default–not immediately, but at some point in the future.  This highly irrational fear would initially manifest itself as a refusal to hold long-dated U.S. treasury securities.  The yields on those securities would therefore rise.  But who cares?  The U.S. government does not need to borrow at the long-end of the curve.  It can borrow at the short-end–indeed, it should borrow at the short-end, to save itself money.  When it borrows at the long-end, it has to pay term-premium to lenders, compensation for requiring them to commit to the loan for an extended period of time.  Paying this premium would make sense if the commitment were of value to the U.S. government–if it created space that made it easier for the U.S. government to find lenders willing to refinance its debt, when the debt comes due.  But the commitment obviously doesn’t have that value, because the U.S. government doesn’t need to find lenders willing to refinance its debt–it effectively has the power to lend to itself.

If the fear of default were to grow acute, it would manifest itself in the form of a refusal on the part of investors to hold short-dated U.S. treasury securities.  The yields on those securities would therefore rise.  What happens next would depend on whether banks retained confidence in the government’s willingness and ability to make good on the securities.

Suppose that banks were to retain that confidence.  The rise in short-dated treasury yields would then create an immediate arbitrage opportunity for them to exploit.  Recall that the Federal Reserve sets the cost of funding for banks by manipulating the aggregate supply of excess reserves available to be lent between individual banks overnight to meet reserve requirements, and by setting the rate on loans made directly to banks for that purpose, through the discount window.  This cost of funding is essentially the short-term interest rate for the U.S. economy.  If the yield on short-term treasury securities were to spike to yields substantially above that rate, then banks could borrow at the rate, use the borrowed funds to buy short-term treasury securities, and collect the spread as profit. Assuming that the U.S. government were able and willing to make good on the securities, this profit would accrue completely risk-free–with zero credit risk and zero duration risk.

It’s tempting to think that if banks were to try to exploit this arbitrage, stepping in and buying high-yield treasuries, that they would have to curtail their loans to the rest of the private sector, to make regulatory space for the loans that they are effectively making to the government.  But this is wrong.  With respect to regulatory capital ratios, treasury securities have a zero risk weighting.  Banks have regulatory space to borrow to buy them in whatever quantity they want, without having to alter any other aspect of their balance sheets or their operations.

But even if banks were unwilling to hold treasury securities, the U.S. government’s ability to borrow would still remain unconstrained.  For the Federal Reserve could solve the problem by directly purchasing the securities in the market, pushing up on their prices and down on their yeilds.  Surely, banks and investors would be comfortable holding the debt of the U.S. government if they knew that the Federal Reserve was in the secondary market, willing to buy them at any price.

To take the example to the maximum extreme, even if banks and investors had become so afraid that they were uncomfortable buying discounted debt that the Federal Reserve was willing to buy from them in the secondary market at par, Congress could simply modify the Federal Reserve Act to allow the Federal Reserve to buy securities directly from the treasury, lending to the treasury directly, without using the secondary market as a conduit.

The point, then, is this. Governments control the interest rates of currencies they issue. The path may be cumbersome, but if a government owes debt in a currency that it has the power to issue, then it can set the interest rate that it pays on that debt as low as it wants, including at zero, which is the interest rate that it pays when it finances itself by issuing money proper.  A situation in which rising interest rates force a currency-issuing government to default is therefore completely out of the question.

Myth #4 — Large Government Deficits Are Necessarily Inflationary

It’s true that large government deficits, and the large equilibrium government debts that those deficits produce, can lead to inflation.  But, as we will explain in the next section, the path is not direct.

As Hume explained in an earlier quote, inflation requires excessive spending–demand that exhausts the economy’s capacity to supply for it.  Large government deficits can certainly be used to finance excessive spending on the part of the government, and such spending can certainly be inflationary.  But it’s not the deficits themselves that produce the inflation, it’s the excessive spending on the part of the government.

Similarly, large government deficits can be used to finance tax cuts and transfers that increase private sector income, leading to excessive private sector spending and eventual inflation. But again, it’s not the deficits themselves that produce the inflation, it’s the excessive private sector spending.  In cases where deficit-financed tax cuts and government transfers are put in place, but do not lead to excessive private sector spending, either because the proceeds are saved, or because there is an output gap, the result will not be inflation.

In summary, large government deficits, run for indefinite periods of time, can provide free lunches, are sustainable, do not lead to rising interest rates, and are not inherently inflationary. The implication, then, is that a policy of fiscal inflation targeting need only focus only on its target, the inflation rate, and that the quantity of government debt that it leaves behind can be ignored. This implication is true in most cases, but it’s not true in all cases. The debt that the policy accumulates can, in theory, become a problem in the future. It’s importantly that we explain how it can become a problem, so that the risks of the policy are not misunderstood.  The sections that follow are devoted to providing that explanation.

Stock and Flow: The Inflationary Mechanism of Government Debt

To better understand the inflationary dynamics of government debt, we need to make a distinction between “stock” and “flow.”  Stock refers to the amount of something that exists; flow refers to the amount of something that moves in a given period of time.

bees2

Imagine a collection of bees swarming around in a cage.  The number of bees in the cage is a stock quantity.  The number of bees that manage to escape from the cage each hour is a flow quantity.

We define money as “legal tender”–whatever must be accepted, by law, to repay debts, public and private.  We define financial wealth as money, or anything that can be readily converted into money in a market, considered in terms of its current market value, minus whatever liabilities are tied to it.  Because financial wealth can be readily converted into money, we can treat it as the functional equivalent of money.

In an economic context, financial wealth has a stock aspect and a flow aspect. The stock aspect is the total amount of it that exists.  The flow aspect is the total amount of it that circulates in the form of expenditures in a given period of time.  Note that by “expenditure” we mean the exchange of money for real goods and services.  The trading of money for other forms of financial wealth is not included, though such trading may occur as part of the process of an expenditure (e.g., I sell a stock in my portfolio to raise money to buy a car).

When a government spends money on a purchase or project, it creates a one-time flow of spending.  This one-time flow takes place regardless of how the spending is funded or financed.  If the flow is inserted into an economy that does not have available resources from which to supply the added demand, the flow will tend to be inflationary.

By “paying for” the spending through taxation, a government can potentially reduce private sector spending flows, “making room” in the economy for its own spending to occur without inflation.  But not all taxation reduces private sector spending flows equally. When taxes are levied on individuals that have a low marginal propensity to spend, the spending flows of the private sector tend to not be substantially affected.  To quote Hume, the taxed money is money would that have remained “locked up in chests” anyways. Taxing it therefore does not free up resources for the government to use.  It is only when taxes are levied on individuals that have a high marginal propensity to spend that taxation reliably frees up resources and offsets the inflationary effects of government spending.

When a government chooses not to “pay for” its spending through taxes, and instead finances its spending with debt or money creation, a stock effect is added to the one-time flow effect of the spending.  Because the spending is never redeemed in taxes, new forms of financial wealth–debt securities and money proper–end up being permanently added to the system as residuals.  The residuals are left without any individual liabilities to offset them.  Of course, they are offset by the liability of government debt, a liability that the private sector bears collective responsibility for.  But, crucially, individuals in the private sector do not view government debt as their own personal liabilities, and therefore do not count it in tallies of their own personal net worth.  Consequently, the net financial wealth of the private sector, as tallied by the individuals therein, increases.

Now, when you increase the stock of something, you tend to also increase its flow, all else equal.  If you increase the stock of bees in a cage, and you change nothing else, you will tend to also increase the number of bees that fly out of the cage as time passes.  To illustrate, suppose that there are 1,000 bees in a cage.  Suppose further that the statistical probability that a given bee will escape in an hour is 0.2%.  How many bees will escape each hour?  The answer: 1,000 * 0.2% = 2.  If you hold that probability constant, and you double the number of bees in the cage to 2,000, how many bees will escape each hour? The answer: 2,000 * 0.2% = 4.  So we see that with the escape probabilities held constant, doubling the stock doubles the flow.

A similar concept applies to financial wealth.  If the net stock of financial wealth in an economy is increased, and if the probability that a given unit of financial wealth will be spent per unit time remains constant through the change, then the amount of financial wealth that circulates in the form of expenditures per unit time–aggregate demand–will rise.

That’s where the true inflationary effect of government debt accumulation lies. Government debt is an asset of the private sector.  When it is held by the central bank, it takes the form of money in the hands of the private sector (the money that the central bank had to create to buy those securities).  When it is held directly by the private sector, it takes the form of government debt securities.  Thus, when government debt is increased, the private sector gains an asset (money or debt securities) without gaining any liabilities (at least not any that it views as such).  It follows that when government debt is increased, the total net financial wealth of the private sector is increased.  Increases in net financial wealth tend to produce increases in spending, and excessive spending can generate inflation.

In the next two sections, I’m going to introduce two related concepts that will be useful in our efforts to understand the inflationary potential of government debt.  Those two concepts are: wealth velocity and wealth capacity.

Wealth Velocity: A New Equation of Exchange

Readers with a background in economics are likely to be familiar with the equation of exchange:

(4) M * V = P * Q

The equation of exchange translates money stock into money flow, using the ratio between them, money velocity.  Here, M is the money supply (the total stock of money in the economy), V is the money velocity (the percentage of the money stock that is spent–i.e., that flows–in a given year, or equivalently, the probability that a given unit of money will be spent in a given year), P is the price index (the conversion factor between nominal dollars and real things), Q is real output (the flow of real things). Note that P * Q is the total nominal spending in the economy.  The equation tells us, trivially, that the total amount of money in the economy, times the probability that a given unit of money will be spent in a given year, gives the total amount of spending in a given year, on average.

The problem with the equation of exchange is that the stock quantity that has the deepest relationship to the flow of spending is not the total stock of money proper, but the total stock of net financial wealth in the economy–money plus everything that can be easily converted into money, considered in terms of its marketable monetary value, minus the monetary value of all debt obligations.  (Note: for convenience, we have often been omitting the “net” term, using the terms “net financial wealth” and “financial wealth” to mean the same thing: money plus marketable assets minus debts.)

The best way to write the equation of exchange, then, is not in terms of M, the total stock of money in the economy, but in terms of W, the total stock of financial wealth in the economy:

(5) W * V = P * Q

Here, Vw is the wealth velocity–the corollary to the money velocity in the original equation of exchange. We can define it as the percentage of financial wealth in existence that gets spent each year, or equivalently, as the probability that a unit of financial wealth will be spent in a given year.  The equation tells us, again trivially, that the total quantity of financial wealth in the economy, times the probability that a given unit of financial wealth will be spent in a given year, gives the total amount of spending that will occur in a given year, on average.

When financial wealth is injected into the private sector through a government deficit, some or all of the wealth may accumulate idly as savings.  Wealth velocity–Vin the equation–will then go down, fully or partially offsetting the increase in W, the stock of wealth.  That’s how large government debt accumulation can occur over time without inflation. The wealth is continually injected via debt accumulation, but the injections coincide with reductions in the velocity of wealth, such that total spending does not increase by a sufficient amount to exceed the productive capacity of the economy and produce inflation.

The problem, of course, is that conditions in the economy can change over time, such that the wealth velocity increases, reverting to its pre-injection value, or rising to some other value.  The prior stock of wealth that was injected, which had been “quiet” because it was being held idly, may then start circulating, and contribute to an inflation.

Wealth Capacity: An Analogy from Thermodynamics

In thermodynamics, there is the concept of “heat capacity.”  The heat capacity of a substance specifies the amount that its temperature will increase when a given amount of heat is added to it.  If its temperature will only increase by a small amount in response to a given injection of heat, then we say that it has a high heat capacity.  If its temperature will increase by a large amount in response to a given injection of heat, then we say that it has a low heat capacity.

Water has a high capacity.  You can add a relatively large amount of heat to a unit of water, and yet its temperature will not increase by very much.  Iron, in contrast, has a low heat capacity.  If you add the same amount of heat to a unit of iron, its temperature will increase significantly.  Intuitively, we can think of the temperature increase as the heat “overflowing” from the iron, which is a poor store of heat.  The heat does not “overflow” from water to the same extent, because water is a good store of a heat.

If the reader will allow us to be somewhat sloppy, we can extend the thermodynamic concept of “heat capacity” to economics, naming the analogous property “wealth capacity.” At a given interest rate, how much financial wealth can an economy store without overheating?  The answer, which is determined by the wealth velocity that will manifest at that interest rate, and the total productive capacity of the economy, specifies the economy’s wealth capacity.  Importantly, an economy’s wealth capacity is specified as a percentage of its potential GDP, which then incorporates productive capacity into the expression.  So, at a given interest rate, an economy might have a wealth capacity of 50% of potential GDP, or 100% of potential GDP, or 200% of potential GDP, and so on.

If an economy has a wealth capacity that exceeds its current quantity of wealth, then it can hold additional financial wealth, and therefore its government can accumulate debt without inflation occurring.  Conversely, if an economy has a wealth capacity equal to or less than its current quantity of financial wealth, then it will not be able to hold additional financial wealth, and therefore government debt accumulation, which involves the injection of financial wealth, will be inflationary.

To make the same point that we made with wealth velocity, wealth capacity is not static, but changes in response to changing macroeconomic conditions.  The risk, then, is that a large injection of wealth will be made, and the economy will appear to be able to absorb it, without inflation.  But over time, as macroeconomic conditions change, the wealth capacity may fall, such that the prior injection, which pushed up the stock of wealth, contributes to an inflation.  That’s why the question of the appropriate level of government debt in an economy is not simply a question of how much financial wealth it can store right now, but a question of how much financial wealth it will be able to store over the long-term, under the different conditions that it will come to face.

The Factors that Influence Wealth Velocity and Wealth Capacity

Though they are not perfect inverses, wealth velocity and wealth capacity are inversely-related to each other.  All else equal, high wealth velocity is associated with low wealth capacity, and low wealth velocity is associated with high wealth capacity.

The following factors influence an economy’s wealth velocity and wealth capacity:

Confidence.  The value of the money that a country issues rests on the confidence that the money will serve as an effective store of value.  As that confidence breaks down, whether in the face of mismanagement or conflict, individuals will be less willing to hold the money, more inclined to spend it in a rush, invest it in real assets, or transfer it abroad.  Thus, as confidence in a country’s money breaks down, its wealth velocity will tend to rise, and its wealth capacity will tend to fall.  The same is true in the other direction.  As a country’s money becomes more credible, more like a reserve currency that the world trusts as a long-term store of value, individuals will be more willing to hold it.  The country’s wealth velocity will then fall, and its wealth capacity will rise.

Location of Homestead.  Do the owners of a country’s financial wealth–the stock of money and marketable securities contained in the country’s economy–live and spend the bulk of their money in that country?  Or do they live and spend the bulk of their money in some other country?  Are they holding wealth in that country because that country is where their responsibilities–financial and otherwise–are located?  Or are they holding wealth in that country because they see an opportunity for a “hot money” profit?  If the former, there will be a greater willingness on the part of the owners of the economy’s financial wealth to hold it, rather than deploy it into consumption or investment, given that the safety and security that comes with holding it will will actually be relevant to the owners’ lives.  It follows that the economy will have a lower wealth velocity and a higher wealth capacity (h/t to @Chris_Arnade for this insightful observation).

This point is important, so let me give an example.  Suppose that I am a Japanese businessman that owns a significant quantity of financial wealth.  Regardless of what happens to the yen’s exchange rate to other currencies, psychologically, I measure the entire financial world in terms of yen.  I fund my lifestyle with yen, all of my debts are in yen.  Having a substantial quantity of yen in savings therefore provides me with a special type of insurance.  It helps ensure that I stay rich, by my own measure of “richness.”  It helps ensure that I retain the ability to fund my lifestyle, and meet my debts.  The fact that there are many people like me, controlling large amounts of Japanese financial wealth, makes it easier for the Japanese government to inject large quantities of yen and yen-denominated securities into the Japanese economy, and find people willing to hold them rather than spend them, even at low rates of return, so that inflation stays low.

But if I am a wealthy Japanese businessman holding my financial wealth in the form of Uruguayan peso, I do not get the same insurance.  Having a substantial quantity of Uruguayan peso in reserve does not reduce the likelihood that I might one day lose the ability to fund my yen-denominated lifestyle, or pay my yen-denominated debts–at least not to the same extent.  And so if the owners of Uruguay’s financial wealth are all people like me, people speculating from abroad, the Uruguayan government will not be able to inject large quantities of new peso and peso-denominated securities into the economy at low rates of return, and have people like me continue to hold them.  We will want to trade them for higher yielding assets in the Uruguayan economy–that’s why we’re involved with Uruguay in the first place, to get a return.  If the prices of those assets get pushed up to unattractive levels in response to the wealth injection, we–or those wealthy foreigners that end up holding our pesos after we sell them–will opt to create new assets in the Uruguayan economy through investment, rather than hold Uruguayan money without compensation.  The result will be inflationary pressure.

Distribution of Financial Wealth.  An economy’s wealth capacity will tend to be higher, and its wealth velocity lower, if the distribution of financial wealth within it is narrow rather than broad.  The reason is obvious.  As an individual’s level of financial wealth increases, her propensity to put additional financial wealth that she comes upon into consumptive use goes down.  It follows that an economy in which the financial wealth is narrowly distributed among a small number of wealthy people will have a lower propensity to consume additional financial wealth, and therefore a higher financial wealth capacity and lower wealth velocity, than an economy where the financial wealth is broadly distributed among the entire population.

Target of the Financial Wealth Injection.  When a government deficit is used to inject financial wealth into an economy, into whose hands is that wealth injected?  Is it injected into the hands of wealth-lacking people who need it to fund their desired lifestyles, and are eager to spend it?  Or is it injected into the hands of wealth-saturated individuals that do not need it to fund their desired lifestyles, and who not are eager to spend it?  The velocity of the wealth–the speed at which it will circulate–will obviously be higher under the former than the latter.

Tendency of Financial Wealth to Collect in Spots.  When wealth is injected and spent, does it continue to move around the economy in a sustained cycle of spending, or does it eventually collect idly in a certain spot, where it gets hoarded?  If it continues to move around, then the wealth capacity will be low; if it tends to collect idly in a certain spot–e.g., in the corporate sector, where it gets hoarded as profit after the first expenditure–then the wealth capacity will be high.

Consumptiveness.  The consumptiveness of an economy refers to the extent to which the individuals that make up the economy are inclined to consume incremental income rather than hold it idly.  All else equal, an economy with higher consumptiveness will have a higher wealth velocity and a lower wealth capacity.  An economy’s consumptiveness is influenced by a number of factors, to include its history, its culture, its demographics, the prevailing sentiment among its consumers, and so on.

Investment Risk Appetite.  The simple fact that an economy likes to save does not mean that it will have a high wealth capacity.  For there are two ways to save.  One can save by holding money idly (or trading it for existing securities, in which case someone else, the seller of those securities, ends up holding it idly), or one can save by investing it in the creation of new assets–new projects, new technologies, new structures, and so on.  The distinction between saving by holding and saving by investing is important because saving by investing involves spending that puts demand on the economy’s existing labor and capital resources. It can therefore contribute to inflation and economic overheating, even as it increases the economy’s productive capacity.

The term “investment risk appetite” refers to the inclination of an economy to save by investing, rather than by holding.  An economy with high investment risk appetite will have a lower wealth capacity than an economy with low investment risk appetite.  As with consumptiveness, an economy’s investment risk appetite is influenced by a number of factors, to include its history, its culture, its demographics, the sentiment and prevailing outlook of its investors, and so on.

In contrast to consumption, the risk that investment will produce inflation is alleviated by the fact that investment adds new assets, new resources, new productivities that the economy can use to supply the additional consumption demand that will be created.  But investment does not deliver those assets, resources, and productivities immediately–there is a time delay.  Moreover, the investment may not be adequate, or appropriately targeted, to supply the additional consumption demand that will be created.  And so inflation and economic overheating are still possible.

The final factor that influences wealth velocity and wealth capacity is the interest rate, a factor that policymakers have direct control over.  Higher interest rates are associated with lower wealth velocity and higher wealth capacity, and lower interest rates are associated with higher wealth velocity and lower wealth capacity.  We discuss this relationship further in the next section.

Interest Rates: The Economy’s Inflation-Control Lever

If pressed, those that are concerned about the risks of large government debt accumulation will usually accept the point that governments can inject financial wealth–new money and debt securities–into the economy without creating inflation, provided that the recipients of that wealth choose to hold it idly rather than convert it into some kind of spending.  But they point out that the simple possibility that the injected financial wealth could be spent, and produce inflation, is reason enough not to make the injection.  To make the injection would be to give the recipients the power to spend, and therefore the power to consume at the expense of other savers, who would lose out in the resulting inflation.

The problem with this point is that economic participants already have the power to consume at the expense of savers.  They can accelerate their consumption, by spending more of what they earn, or by borrowing.  The acceleration will stimulate inflation, which will occur at the expense of those that have chosen to save.  So the inflation risk introduced by government debt accumulation, and associated private sector wealth injection, is a risk that already exists at the current level of government debt, and that would exist at any level of government debt.

For an economy that contains a given quantity of financial wealth, the amount of inflation that it experiences will be determined by the balance of “holding money idly” vs. “deploying money into consumption and investment” that takes place within it.  That balance can shift at any time.  Fortunately, governments have a tool that they can use to manage the balance, so as to control inflation.  That tool is the interest rate.  The interest rate is the expense that individuals must pay in order to borrow money to consume and invest. Conversely, the interest rate is the reward that individuals receive in exchange for holding money idly, rather than deploying it into consumption or investment.  When individuals hold money idly, they take it out of circulation, where it cannot contribute to inflation.

When an economy is suffering from too much activity, the government will set the interest rate at a high level, providing generous compensation to whoever willingly agrees to hold money idly.  When an economy is suffering from too little activity, the government will set the interest rate a low level, removing the reward–or worse, imposing a punishment–on whoever chooses to hold money idly.

In terms of the risks of large government debt accumulation, the primary risk is that the debt will make it more difficult for policymakers to change interest rates in response to changing economic conditions–changing wealth velocities and wealth capacities.  When interest rates are increased in the presence of an overwhelmingly large government debt, the interest expense that the government incurs on that debt increases.  If the increased expense is funded with tax increases and spending cuts, the economy will suffer a destabilizing effect, both economically and politically.  If the increase is funded with additional debt, the result will be added inflationary pressure, because government debt accumulation entails additional private sector wealth injection.

To come back to the example of Japan, right now, Japan is underutilizing its labor and capital resources, a fact demonstrated by its 0% inflation rate.  In terms of stimulus, Japan would benefit from a large fiscal deficit, run over the next several years, if not for longer. But conditions in the Japanese economy might one day change, to where the population’s propensity to consume and invest, rather than hold savings idle, meaningfully increases. If that propensity does change, and if Japan has an enormous government debt to finance when it does, then controlling inflation with interest rate increases could become difficult, if not impossible. I explore this scenario in a later section.

Asset Prices: A Second Inflation-Control Lever

It turns out that there is an additional channel through which interest rates can influence inflation.  Interest rates–in specific, the interest rate paid on cash (money and very short-term low-risk debt securities)–affects the prices of all existing assets.  Rising interest rates make the return on cash more competitive with the return on existing assets, and therefore tend to cause the prices of those assets, in cash terms, to fall.  Conversely, falling interest rates make the return on cash less competitive with the return on existing assets, and therefore tend to cause the prices of those assets, in cash terms, to rise.

Crucially, the quantity of financial wealth in an economy includes not just money proper, but all forms of financial wealth contained within it–debt securities, equity securities, real estate, collectibles, and so on.  When the prices of these assets rise, the stock of financial wealth in the economy rises–not only in a gross sense, but in a net sense, because the liabilities that the assets match to don’t increase in the same way.  The increase in the stock of financial wealth has the potential to lead to an increase in its flow–spending.

The pass-through from asset values to spending tends to be weak with respect to debt and equity securities, because those securities tend to be owned by small, wealthy segments of the population that have a low marginal propensity to spend, and because changes in the prices of the securities aren’t interpreted or trusted to be permanent. But in asset markets where ownership is more evenly distributed across the population, and where the upward price trend is interpreted to be more stable and reliable–in housing markets, for example–the pass-through can be significant.

Trapped Monetary Policy: How Things Can Go Wrong

The best way to illustrate the risk of large government debt accumulation is to use an extreme example.  So here we go.  Suppose that Japan has implemented our recommended policy of 2% fiscal inflation targeting. Suppose further that conditions in Japan are such that to maintain the economy on a 2% inflation target, the country needs to run a 20% deficit with interest rates at 0%.  Suppose finally that the long-term nominal growth rate under the given conditions will be 3%. Assuming that Japan starts with net government debt at its current value, roughly 134% of GDP, the debt would rise to roughly 500% by 2055 and 667% at equilibrium (20% / 3%), to be reached well over 100 years from now.

Now, let’s fast forward to 2055.  By then, conditions in the Japanese economy may have changed.  After 40 years of sustained 2% inflation, the cultural propensity of Japanese consumers to spend rather than save, and of Japanese savers to invest in the real economy rather than hold yen idle, may have increased.  The demographic profile may have improved.  Much of the financial wealth injected by the deficits may have “leaked”, through channels of consumption, investment, inheritance, and charity, from high net worth segments of the economy, where it was pooling, to lower net worth segments of the economy, where the marginal propensity to spend it will be higher.

In practice, these changes, if they were to occur, would be expected to occur gradually, allowing the central bank time to adjust fiscal and monetary policy so as to accommodate them.  But to make the dynamic vivid, let’s assume, for the sake of argument, that the changes occur instantaneously, in the year 2055.  To summarize the scenario, Japan runs a 20% deficit for 40 years, accumulates a government debt worth 500% of GDP, and then suddenly, the macroeconomic backdrop changes.

To frame the dynamic in terms of wealth velocity, the changes, when they occur, will cause the wealth velocity in Japan to increase significantly.  The significant increase in wealth velocity will apply to a very large stock of financial wealth–500% of GDP–and will likely push the economy’s total spending to levels that will exceed its available labor and capital resources.  The result will be inflationary pressure.  To frame the dynamic in terms of wealth capacity, the changes, when they occur, will significantly reduce Japan’s capacity to store financial wealth. The country will no longer be able to hold money and government debt securities worth 500% of GDP at zero interest rates.  The result, again, will be inflationary pressure.

How will the government deal with this pressure?  Clearly, it will need to start by raising taxes and cutting spending, so as to bring its enormous 20% deficit down to something small.  But that will only remove future injections of financial wealth; it will not reverse the prior injections, the potential circulation of which represents the primary inflationary threat.  To reverse the prior injections, the government would have to run a surplus, and there is no way that surpluses sufficient to unwind more than 300% of GDP in added financial wealth could be run in any reasonable amount of time.

Ultimately, the only way to prevent the large stock of wealth from circulating and stirring up inflation would be through an increase in the interest rate.  By increasing the interest rate, the BOJ would give Japanese wealth owners the needed incentive to hold the large supply of yen cash and debt securities that will need to be held.  Without that incentive, the supply will be tossed around like a hot potato, fueling an inflation. (Note: the prices of the debt securities would fall in response to the interest rate increase, causing their yields to rise and making them more attractive to hold, and also lowering the total market value of financial wealth in the economy).

Let’s suppose that in order to manage the inflationary pressure, the BOJ would need to raise interest rates from 0% to 7%.  At an interest rate of 7% and a total debt stock of 500%, the government’s added debt expense would amount to 500% *  7% = 35% of GDP. That’s an enormous expense.  Where would the money to fund it come from?

The government could try to increase taxes or cut spending to make room to pay it, but the total size of the tax increases and spending cuts that would be necessary would be 35% of GDP–a number that would be extremely difficult to successfully draw in, and highly economically and politically destabilizing, if a way to draw it in were found.

It’s likely, then, that the Japanese government would have to borrow to pay the interest. But borrowing–by running a deficit–would entail the injection of a substantial quantity of new financial wealth–35% of GDP, if all of the interest were borrowed–into the already overheating private sector.  That injection–which would accrue to the holders of yen cash and debt securities in the form of  interest income–would represent an additional source of potentially unmanageable inflationary pressure, undermining the effort.

Now, to be fair, it’s possible that Japan could find some way to combine tax increases, spending cuts, and deficit borrowing to pay the interest expense.  But we can’t ignore the inflationary effect of the market’s likely behavioral response to the difficulty.  In practice, economic participants observing the country’s struggles would become increasingly averse to holding yen cash and yen-denominated debt securities.  This aversion would eventually become reflexively self-fulfilling.  Wealth velocity would then rise further, with wealth capacity falling further, adding further inflationary pressure.  Investors would seek to transfer wealth abroad, causing the yen to depreciate relative to other currencies. The depreciation would make Japanese goods more attractive to the world, again, adding further inflationary pressure.

We can see, then, how an eventual bout of high inflation might become unavoidable, even as the central bank does everything in its power to stop it.  The central bank’s preferred tool for combating inflation–the interest rate–would effectively be trapped by the enormous government debt.  Its use in the presence of that debt would exacerbate the inflation, or destabilize the economy, or both.

This type of scenario is not purely theoretically, but has actually played out in economic history, most notably in France in the 1920s.  In more benign contexts, there have been a number of examples of central banks that were forced to surrender to inflation, prevented by their heavily indebted sovereigns from setting interest rates at the levels needed to maintain control over it.  A well-known example is the United States in the years shortly after World War 2, where the Treasury effectively forced the Federal Reserve to hold interest rates at zero, despite double-digit inflation.

In each of these cases, the inflation proved to be a short, temporary occurrence, rather than an entrenched, long-lasting phenomenon.  Conveniently, the problem of the large government debt ended up correcting itself, because the inflation it provoked substantially reduced its value, both in real terms and relative to the nominal size of the economy and the tax base.  The outcome may have been unfair to those savers who ended up holding money at deeply negative real interest rates, but it did not entail any larger humanitarian harms.

Essentially all of the known historical cases in which large government debt has led to inflation have involved war.  One reason why we might expect that to be the case is that war temporarily disables substantial portions of the labor and capital stock of an economy, reducing its productive capacity.  But there’s an additional reason.  War is expensive, it requires the government to take on substantial quantities of debt.  Those quantities are taken on to fund an urgent activity, and so they are allowed to accumulate even as they contribute to excessive inflation.  Crucially, when the war ends, the accumulated stock of debt does not go away.  It is left as financial wealth in the private sector.  As people return home and begin life again, that wealth starts circulating in an inflationary manner–sometimes quickly.

Unlike in war, where government debt is taken on to fund an urgent activity, in fiscal inflation targeting, government debt is taken on in an effort to keep inflation on target, given structural weakness in aggregate demand.  Policymakers therefore have space to respond to early signs of rising inflationary pressure–signs that the structural weakness is abating.  If Japan were to implement fiscal inflation targeting, and needed to run a deficit worth 20% of GDP with rates at zero to achieve that target, the scenario would not play out as Japan running that deficit for 40 years without seeing any changes, and then suddenly seeing its wealth velocity rise dramatically and its wealth capacity fall dramatically.  Rather, the country would run high deficits, macroeconomic conditions would gradually change, with wealth velocity gradually rising and wealth capacity gradually falling, the changes would show up in real-time measurements of inflation, and the central bank would respond, adjusting its fiscal stance, not in an emergency after an unmanageable debt has already been accumulated, but gradually, as the process moves along.

Fiscal Inflation Targeting: A Cost-Benefit Analysis

With the potential inflationary risk of large government debt accumulation now specified–inflation via the mechanism explained in the prior section–we are in a position to weigh the costs of fiscal inflation targeting versus its benefits, and to outline the characteristics that would make an economy with weak aggregate demand and abnormally low inflation a good candidate for the policy.

First, the benefits.  Unlike monetary manifestations of the same concept, fiscal inflation targeting would actually work.  No more of the frustration of having to watch inflation consistently undershoot the central bank’s target (core inflation is below 2% in almost every developed economy in the world right now), and having to bear with the central bank’s excuses for why that’s not a problem, its reasons for remaining optimistic that its targets will eventually be reached.  The central bank would no longer be limited to the use of a placebo, but would have an incredibly powerful tool at its disposal, the ability to inject financial wealth directly into the economy, targeting those segments of the economy where the injections would have the greatest effect.  This tool would give it the ability to forcefully defend its targets, an ability that it does not currently have.

Importantly, the central bank would be able to craft the injections around basic principles of fairness, distributing them broadly and evenly, so that everyone gets to partake in their benefits.  Contrast that with the current monetary approach to stimulus, quantitative easing, which does little more than inflate asset prices through a psychological effect, increasing the financial wealth of only those people who already own it, the people that are the least likely to need a boost.

The ability would allow the central bank to more efficiently use monetary policy for other purposes, such as the promotion of financial stability.  A central bank that had concerns about the financial stability risks of low interest rate policies–increases in private sector leverage, the emergence of asset bubbles, and so on–could run a tighter monetary policy alongside a looser fiscal policy, mitigating financial stability risks without causing the economy’s performance to fall below its potential.

The injections would not directly occupy any resources, as wasteful government spending might, but would instead leave it to the private sector to determine how resources are allocated.  Conducting the stimulus in this way would lead to more efficient labor and capital formation, maximizing the economy’s output over the long-run.

Now, the cost.  The cost is the risk that a debt will build up over time that is so large that it will obstruct the government’s ability to use monetary policy to control inflation as needed.  In the context of inflation targeting, this cost is mitigated by two factors:

  • The debt is being accumulated in response to structural economic weakness, rather than in response to some temporary urgency, such as war.  The likelihood of a rapid inflationary change in macroeconomic conditions–the kind of change, for example, that would occur when a war ends and when everyone returns home to start normal life again–is low.  If the economy’s wealth velocity and wealth capacity do change over time, they will tend to change gradually, affording the central bank space to respond by withdrawing its fiscal injections, or even reversing them by running surpluses. Crucially, with the central bank in control of fiscal policy, the political obstacles that usually get in the way of an appropriate response will not apply.  The central bank will be free to do whatever is needed, without having to worry about the impact on the next election.
  • The cost, even if it is incurred, is a brief, self-correcting event.  If the government’s debt prevents the central bank from raising interest rates as needed to control inflation, and if cyclical conditions improve, then the economy will simply have to endure a period of high inflation.  Conveniently, the period will reduce the value of the debt in real terms, and also relative to the nominal size of the economy and the tax base, which will substantially increase.

The cost-benefit analysis of fiscal inflation targeting is most attractive in the following type of economy:

  • An economy where the currency is issued by a government with a history of political and economic stability, backed by a disciplined and credible central bank.  Ideally, an economy that controls a reserve currency that the rest of the world seeks to save wealth in the form of.
  • An economy whose stock of money and debt securities is held mostly in domestic hands, or in the hands of foreign individuals that get special insurance from holding the securities in lieu of the securities issued by their own countries.
  •  An economy with a high level of wealth inequality, where wealth exhibits a tendency to pool in certain concentrated locations–for example, the corporate sector–when it tries to circulate.
  • An economy with low consumptiveness and low investment risk-appetite.
  • Most importantly, an economy with the above characteristics, where the characteristics have structural explanations, and are expected to remain in place for the long-term.

An economy that meets these criteria will have a high wealth capacity, and will retain that capacity over time, allowing for fiscal injections to be made that stimulate the economy and bring it to its inflation target over the short-term, without incurring undue risk of a larger inflation problem over the long-term.  Conveniently, the mature, developed, aging countries of the world that have the greatest need for ongoing fiscal stimulus–the U.S., Japan, the U.K., and the countries of the Eurozone, especially the creditor countries–exhibit most or all of these characteristics, and are therefore excellent candidates for the policy.

It’s impossible to conclusively know how large an economy’s wealth capacity is until it is reached.  And so, for all we know, a country like Japan might have the ability to hold money and debt securities worth 1000% of its potential GDP, or 2000%, or 3000%, or higher, with interest rates exactly where they are now–at zero–and not overheat.  The fact that the country has accumulated so much debt over time, and yet still struggles to stay out of deflation, suggests that the value is very high, much higher than economists assume, and definitely much higher than its current net debt of 134% of GDP.

The right approach for Japan, then, is to push the limits and find out.  The risk is an unlikely trapping of monetary policy that leads to a one-time inflation that eventually corrects itself; the reward is a likely discovery of a free lunch that changes the economic understanding of government debt forever, and that paves the way for a much-needed solution to the looming problem of demographic and secular stagnation, not only for Japan, but for the entire world that will eventually have to confront it.

Posted in Uncategorized | Comments Off on Fiscal Inflation Targeting and the Cost of Large Government Debt Accumulation

The Trajectory of a Crash

It’s amazing to think that just last Monday, August 17th, the S&P 500 closed at 2102.  Today, it closed at 1868, falling 11.1% in 6 trading days.  The shocking speed of the decline has injected a level of fear into markets not seen since the fall of 2011, when the Eurozone debt crisis was reaching its apex.  Many traders have referenced 1987 as a paradigm for what might happen in a worst case scenario over the coming days and weeks, so I figured it would be interesting to explore where exactly a 1987 scenario would take us in terms of prices.

The chart below shows the hypothetical price trajectory of the S&P 500 over the next 2 years if the current market ends up performing an exact repeat of the 1987 crash, with Friday, October 2nd, 1987 as the crash starting point:

2yr1987

The closing low will occur on Monday, October 19th, 2015–less than two months from now–at an S&P level of 1435, a 32% correction in full.  Fortunately, investors that hold their positions through the plunge will get their money back in short order, less than two years.

The next chart shows the hypothetical price trajectory of the S&P 500 over the next 2 years if the current market ends up performing an exact repeat of the 1974 crash, with Wednesday, March 13th, 1974 as the starting point:

2yr1974

The closing low will occur on Tuesday, March 8, 2016, at an S&P level of 1311, completing a 37% correction in full.  Again, investors that hold their positions through the plunge will get all of their money back within two years.  But that’s only true in nominal terms.  To get their money back in real, inflation-adjusted terms, with reinvested dividends included, they will have to wait until March of 2022.

The next chart shows the hypothetical price trajectory of the S&P 500 over the next 3 years if the current market ends up performing an exact repeat of the 1937 crash, with Wednesday, August 25th, 1937 as the starting point:

3yr1937

The closing low will occur on Tuesday, March 22nd, 2016, at an S&P level of 1144, for a 46% correction in full.  In nominal terms, with reinvested dividends included, investors will have to wait until March of 2021 to get their money back.  In real terms, they won’t get their money back until January of 2023.

Finally, the big one.  This last chart shows the hypothetical price trajectory of the S&P 500 over the next 3 years if the current market ends up performing an exact repeat of the 1929 crash, with Thursday, October 17th, 1929 as the starting point:

3yr1929

The closing low will occur on Tuesday, May 8th, 2018, at an S&P level of … brace for it … 253, an 88% correction in full.  On a nominal total return basis, investors will not get their money back until October of 2030.  Interestingly, given the severe deflation of the period, that date will end up coming much sooner in real terms–October of 2022.

Personally, I don’t expect a correction commensurate with any of these scenarios to play out. Even a 20% correction would surprise me.  But history teaches us that large downward price moves are, and always have been, real possibilities in a market, even when everyone has a story for why they are unlikely.

Posted in Uncategorized | Comments Off on The Trajectory of a Crash

Profit Margins in a “Winner Take All” Economy

The following chart shows the aggregate net profit margin of the S&P 500 using earnings data updated through the 1st quarter of 2015 (75% complete):

spx net profit margin

With yet another quarter now on the books in which profit margins have remained steady at record highs, it’s becoming increasingly difficult for open-minded investors to reject the possibility that “this time is different”–i.e., the possibility that the observed profit margin increase relative to past averages is secular in nature, and that the mean reversion that many have been expecting simply isn’t going to happen.

If the profit margin increase is secular, what is driving it?  Analysts who write on the topic tend to cite two factors associated with the cost structure of the corporate sector: (1) weak labor bargaining power leading to reduced labor costs, and (2) low interest rates leading to reduced interest expense.  Because these factors are likely to remain in place going forward, analysts have argued that profit margins will remain elevated.

On the labor front, labor bargaining power has weakened substantially amid globalization, automation, and the demise of unions.  As a percentage of final sales, labor costs–which consist primarily of wage and pension expenses–have fallen as a percentage of final sales from a previous long-term range of 62% to 64% to a new low of roughly 56%, which translates to a profit margin boost of roughly 7%.  From BEA NIPA Table 1.14:

ae3a

On the interest front, the picture is quite different.  Interest rates have fallen by more than 10% over the last 30 years, but interest expense hasn’t shown a proportionate drop.  As a percentage of final sales, the reduction in interest expense has amounted to a meager 2.5%.  Interestingly, current corporate interest expense is almost twice as high as it was in the 1960s, despite the fact that long-term corporate bond yields are lower today than they were then.  Again, from NIPA Table 1.14:

zeze

The reason that interest expense hasn’t fallen on par with the fall in interest rates is that corporate debt levels have grown substantially alongside the fall.  Recall that total interest expense depends not only on the interest rate paid, but also on the quantity of debt that the interest must be paid on.

The following complicated chart demonstrates the point graphically.  The blue line is the interest expense of nonfinancial corporations as a percentage of final sales.  The orange line is an approximation of that expense using nonfinancial corporate debt outstanding and Moody’s BAA yield as functional inputs.

dfae

In terms of the explanations themselves, reductions in various components of the corporate cost structure may have helped to catalyze the profit margin increase, but they do not explain how corporations have been able to hold on to profit margins in the face of competition.

On classical capitalist economic theory, reductions in corporate costs should not lead to sustained increases in profit margins.  The reason is simple.  With all else equal, increased profit margins imply increased returns on (newly) invested capital (ROIC).  Increased ROICs tend to provoke increased investment inflows into the sector or industry where the ROICs have increased.  These inflows add capacity and therefore intensify competition. They also put upward pressure on corporate costs–wages and interest rates.  The combined result is downward pressure on ROICs and therefore downward pressure on profit margins.  To quote the inventor of capitalism himself:

“The increase of stock, which raises wages, tends to lower profit. When the stocks of many rich merchants are turned into the same trade, their mutual competition naturally tends to lower its profit; and when there is a like increase of stock in all the different trades carried on in the same society, the same competition must produce the same effect in them all.” — Adam Smith, The Wealth of Nations, 1776, I.IX.2

It’s clear, then, that explanations that point to lower labor costs and lower interest expense cannot be the whole story.  They fail to explain how the mechanism of mean-reversion in profitability, a fundamental tenet of the way capitalist economies operate, could be circumvented for so long.

Ultimately, on classical economic theory, there are only two ways for profit margins to experience sustained increases over the long-term:

First, corporate agents can become more risk-averse.  If they become more risk-averse, then higher ROICs will be necessary to entice them to invest–that is, the reward to investment will have to increase to get them to come off the sidelines and take risk.  Investment is the basis for competition, and therefore if higher ROICs are necessary to entice corporate agents to invest, then higher ROICs will be necessary to entice them to compete with each other.  The result will be higher ROICs at the eventual competitive equilibrium.

The data, however, do not support the claim that risk-aversion has increased.  Corporate investment as a percentage of GDP, for example, is higher now than it was at the prior cycle high, and at roughly the same level as the highs of the cycles of the 1960s and early 1970s.

corpinv

A second mechanism through which ROICs can sustainably increase is through increases in barriers to entry.  Barriers to entry keep competition out.  As they get stronger, the players protected by them are able to successfully operate at increased levels of profitability, free from the threat of competition.

This brings us to the subject of the piece, the “Winner Take All” economy.  The point is difficult to quantify or conclusively prove, but it seems that the dramatic technological changes of the last 20 years have made credible competition in certain key sectors of our economy more difficult, and have allowed dominant best-in-breed companies–the $AAPLs, $GOOGs, $MSFTs, $FBs, and so on of the world–to command sustainably higher profit margins.

The current U.S. economy seems to have more genuine monopolies than the economies of old–more companies that face little to no competition.  The increase in monopoly businesses and monopoly products seems be due, at least in part, to the massive distributional, network-creative and network-protective power of the internet, and also the shift towards the production of non-physical things.  A first-mover with a strong intangible product can distribute that product to the entire world at little cost, protect it as intellectual property, and build a profitable user network around it that other corporations will have an increasingly difficult time competing with.

Think for a moment: how would one go about competing with the likes of an $AAPL, $GOOG, $MSFT or $FB–the Iphone, Google search, Windows/Office, or Facebook? Would it even be possible?  These companies have tried to enter into each other’s domains in the past, but they’ve never succeeded–in fact, they’ve never even come close.  Every competitive effort has turned out to be a hopeless waste of time and money–a Microsoft phone, a Facebook search, a Google Plus, and so on.  It’s no wonder, then, that these companies have been able to enjoy elevated profit margins–in excess of 20% on a net basis–that would have been unheard of 50 years ago.  The effect seems to extend, albeit to a lesser degree, to dominant non-technological companies that have been able to leverage modern technology to efficiently expand their customer bases, the pervasiveness and relevance of their brands, and the dominance of their market positions.

To be fair, there are a number of other, non-technological explanations that one can point to in to account for the increase in barriers to entry.  Our economy, for example, has become increasingly complex from a regulatory perspective, and complex regulation tends to make new entry more difficult.  Additionally, over the last few decades, the corporate sector has shown an increased preference for deploying excess capital into mergers and acquisitions rather than new investment, which naturally tends to reduce competition.  The point, however, is that technology, with its creation of massive, ubiquitous companies that are literally impossible to compete with, appears to be the single biggest driver of the profit margin increase in aggregate.

The possibility that increased barriers to entry represent the primary causal factor behind the observed profit margin increase is supported by the fact that the increase has not been broad-based, as would be expected if it were a simple consequence of weak labor bargaining power, low interest rates, or some other generic factor associated with the corporate cost structure.  Rather, it is concentrated in specific sectors–especially the technology sector–and in specifically dominant individual large cap, blue-chip names.

Fortunately, we don’t have to accept the explanation on faith.  We can test it empirically, in actual data.  If a “Winner Take All” economy, with its associated barriers to entry–first-mover barriers, network barriers, patent barriers, size barriers, regulatory barriers, and so on–has allowed an increasingly concentrated group of dominant companies to earn substantially higher profit margins, then we should expect the following.  If we separate companies in the market into different tiers based on their profit margins, we should expect the higher tiers to have seen larger increases in their profit margins in recent decades than the lower tiers.  Increases in the profit margins of the higher tiers–at the expense of the lower tiers–would represent the “Winner” gradually “Taking All.”

I recently asked my favorite blogger, the brilliant Patrick O’Shaughnessy of O’Shaugnessy Asset Management (Twitter: @millennial_inv, Blog: http://www.investorfieldguide.com), to put the hypothesis to the test by separating companies in the market into different bins by profit margin, and then charting the aggregate profit margins of each bin.  If our explanation is correct, then the aggregate profit margins of the higher bins should have increased more over the last few decades than the aggregated profit margins of the lower bins.  Lo and behold, that’s exactly what the data shows.  The profit margin increase of the last 20 years has not been broad-based, shared by all bins, but has instead been concentrated in the highest profit-margin bins.  The companies in those bins have seen their profit margins explode, while the companies in the lower-tier bins have seen little if any increase–and for some bins, an outright reduction.

The following chart, taken from Patrick’s recent blog post, “The Rich Get Richer”, beautifully illustrates the point:

richgetricher

Notice that the profit margins of the various bins moved roughly commensurately up until the late 1980s and early 1990s.  At that point, something happened.  The profit margins of the top bin proceeded to explode, rising by over 1000 basis points (bps).  The profit margins of the next two highest bins stayed roughly flat. And the profit margins of the two lowest bins actually fell–even as the labor and interest costs of companies in those bins were supposedly reduced.  Overall, the dispersion of profit margins increased dramatically–which is the hallmark sign of a “Winner Take All” economy.

The effect was particularly pronounced in the two sectors that we wrote about previously–technology and finance–which together make up more than 40% of S&P 500 earnings.

fintech

As the chart shows, the profit margins of the top bins in these sectors have increased by over 1000 bps.  The profit margins of the lowest two bins, in contrast, have either gone nowhere  or outright contracted.

For reference, here are individual graphs for all sectors:

cat1

cat2

cat3

As our account would predict, companies that are farther away from the “new economy”–for example, companies in the energy and material sectors that sell basic commodities and that are price takers, or companies in the utility sector whose profit margins are determined by the government–do not show the effect.  For the most part, the high profit margin bins in those sectors haven’t seen any more of a profit margin increase than the other bins.  The effect is instead concentrated in sectors where the dominant players have pricing power–technology, finance, consumer staples, consumer discretionary, health care, and to a lesser extent, industrials.

It remains an unresolved question as to whether and for how long the trend towards a “Winner Take All” economy will persist.  But open-minded investors should admit that it could persist for a very long time, if not forever–and that it could even extend further, with profit margins rising further.  At any rate, given that profit margins have stayed firmly elevated for such a long time without any signs of sustainably falling, and given that we now have a compelling explanation for why they would be expected to stay elevated over the long-term, bearishly-inclined investors should seriously consider the possibility that the “mean-reversion” that they’ve been patiently waiting for isn’t going to happen, at least not to the extent expected.

Posted in Uncategorized | Comments Off on Profit Margins in a “Winner Take All” Economy

Capital Recycling at Elevated Valuations: A Historical Simulation

Those who expect U.S. equities to deliver poor returns going forward can cite two compelling reasons in defense of their expectation:

(1) Equity prices are significantly elevated relative to underlying earnings fundamentals.  The S&P 500’s trailing price-to-earnings ratio, for example, is 20.5 on a GAAP basis and 18.8 on an operating basis, more than a full standard deviation above the historical average of ~14 for GAAP, and ~13 for operating.

(2) Earnings, which make up the E in the P/E  ratio, are artificially high, having been pushed up by elevated levels of corporate profitability, which are anywhere from 30% to 70% above their historical averages, depending on the choice of measurement.

Removing the effects of changes in valuation, the average historical real total return for U.S. equities has been roughly 6% per year.  If U.S. equity P/E ratios and profitability levels were to fall back to their historical averages, this 6% return would get dragged down to roughly zero.  Over a 10 year horizon, the P/E ratio compression would subtract roughly (13/18.8)^(1/10)-1 = 3.6%.  The profitability compression, on the generous assumption that current profitability is only 30% above its natural level, would subtract another (1/1.3)^(1/10)-1 = 2.6%.

The problem, of course, is that it’s possible that “this time is different”–with respect to both P/E ratios and profitability levels.  It’s true that P/E ratios are substantially elevated relative to the long-term historical average, but that average might not represent the natural level for the current environment, which is characterized by:

  • Aggressive policymaker advocacy and monetary support for equity markets, rendered possible by an environment of persistently weak inflation.  This advocacy and support increases investor confidence and creates an environment in which there is no alternative (T.I.N.A) to equities for anyone that wants to earn a return (which is almost everyone).
  • Greater cultural affinity for equity investing, brought about, in part, by the historical lesson, now learned by virtually all, that equities are the best place to invest money for the long-term. Equities just don’t get cheap like they used to.  The market in general is too efficient, too adapted, too familiar with its own history to allow that to happen.

These arguments are sure to raise the hairs on certain people’s skins–“this time is different” is a very dangerous claim.  But clichés aside, the underlying claim might be true.  The fact that valuations have managed to stay historically elevated for so long–well over 20 years now–without showing any sign of retreating, increases the probability that the claim actually is true.

On the profitability front, the U.S. economy has evolved dramatically over the course of history.  It’s quite possible that in this evolution, barriers to competitive entry have emerged that didn’t previously exist–first-mover barriers, network ownership barriers, regulatory barriers, patent barriers, and so on.  In trying to understand how dominant, best-in-breed companies–the $MSFT’s, $GOOG’s, $FB’s, and $AAPL’s of the world–have been able to to capture and hold on to absurdly high levels of profitability, in contravention of normal competitive forces, these barriers would seem to be an obvious culprit.

The fact that profitability levels have stayed high for over 20 years now, showing little inclination to sustainably retreat, gives support to this view.  Additional support is provided by the fact that the elevated profitability levels seem to be concentrated in very specific industries and capitalization categories–technology, finance, and large cap multinational–rather than evenly distributed across the overall market.  For that reason, there’s likely to be a sustainable causal explanation for their emergence.

Now, none of this is offered to suggest that valuations and profit margins won’t retreat going forward.  My own view is that they will retreat, and are already in the process of doing so. I just don’t think that they are likely to retreat all the way back to past averages. It seems to me that to expect such an outcome, one has to completely ignore the relevant differences that exist between the modern era and prior eras.

One thing we can be reasonably sure of, however, is that if U.S. equity valuations stay where they are for the long-term, then returns will suffer by a different mechanism.  That mechanism is what I’m now going to try to quantify.

Recall that the Total Return EPS index is an index that tells us what EPS would have been if all historical dividends that were actually paid out to shareholders had instead been diverted into share buybacks, where they would have stayed inside the equity.  Crucially, in constructing the Total Return EPS index, we conduct the share buybacks at hypothetical prices corresponding to the same valuation across history, rather than at the prices that were actually quoted in the market, which encompassed significantly different valuations at different points in time.

Because share buybacks are functionally identical to reinvested dividends in terms of their effects on total return, and because the Total Return EPS index assumes that all share buybacks are conducted at the same valuation, the index effectively tells us what the Total Return to investors would have been if the effect of changes in valuation had been completely removed.

To illustrate, suppose that the U.S. equity market had always traded at 19 times earnings.  If we had bought the market in January 1966, at 19 times earnings, reinvested all of our dividends at 19 times earnings, and then sold out 10 years later, in January 1976, at 19 times earnings, what would our return have been?  To get to the answer, we simply build a Total Return EPS index on the assumption that all share buybacks are conducted at 19 times earnings.  We then calculate the annualized rate of growth of that index from January 1966 to January 1976.  That rate of growth will be the hypothetical total return under the stipulated conditions of constant valuation.

The market’s P/E ratio is substantially elevated on a historical basis.  In order for this elevation to not impose an eventual drag on returns, the elevation will have to persist indefinitely into the future.  But if it persists, then the dividends that corporations pay out to shareholders will be reinvested at historically expensive prices.  Because those dividends will be expensively reinvested, they will accrete at historically depressed rates of return, producing historically depressed total returns.  There is essentially no way to escape this conclusion.

For perspective, if, from 1871 to 2015, the U.S. stock market had always traded at its average historical valuation, shareholders would have earned an average rolling 10 year real total return of approximately 6.3% per year.  Of that return, 1.65% would have come from organic EPS growth, and the other 4.65% would have come from dividends reinvested into the market. But if those dividends had not been reinvested, and had instead been kept idle in a brokerage account collecting interest at the short-term rate, their return contribution would have fallen from 4.65% to well less than 1%, for a total return below 3%, less than half of the actual.  Right off the bat, then, we can appreciate the fact that the reinvestment of dividends, and the implied rate of return at which the reinvestment is conducted, matters to total return–big time.

To quantify the impact that varying dividend reinvestment valuations have on total return, I’m now going to run a series of simulations.  I’m going to build 25 different Total Return EPS indices for the period January 1871 to March 2015, with the share buybacks in each index conducted at Shiller CAPE levels ranging from 5 to 30 in unit increments.  For each Total Return EPS index, I’m going to calculate the average rolling 10 year real growth rate of the index across its history, which, you will recall, is the average rolling 10 year real total return that investors would have earned if valuations had stayed constant at the specified Shiller CAPE level.  I’m then going to chart the total returns as a function of the different valuations. The ensuing charts will give us a clear picture of how much total return we should expect to lose if valuations stay elevated at present levels.

Before I can do that, I need a non-controversial formulation of the Shiller CAPE.  To that end, I’m going to use the Total Return EPS CAPE (the Shiller CAPE adjusted for changes in dividend payout ratios).  I’m going to employ two specific versions of that CAPE: one built on GAAP earnings (which include the questionable goodwill writedowns of the post-2002 era, writedowns that resulted from the application of accounting standards that were not applied to prior eras and that therefore make for distorted historical comparisons), and one built with operating earnings substituted in after 1998 (which exclude the goodwill writedowns, but which might also exclude other types of justified accounting losses that would make a historical comparison to GAAP earnings unfair).

The two versions are shown below alongside the original Shiller CAPE.

all3comp

As you can see, the Total Return EPS CAPE built on GAAP earnings (yellow, current value: 25.96) is not much different from the original Shiller CAPE (red, current value 27.52), which suggests that historical changes in dividend payout ratios haven’t appreciably affected the accuracy of the original.  However, the post-2002 writedowns make a big difference.  When removed, they bring the Total Return EPS CAPE (blue) down to a current value of 21.93.  Note that all of these CAPEs have been normalized so that their averages are equal to the average of the Original Shiller CAPE–14.19 on a harmonic basis.

The Total Return EPS CAPE built on GAAP earnings can serve as a reasonable upper bound for valuation.  It’s unlikely that the market is more expensive than indicated by that measure.  Similarly, the Total Return EPS CAPE built on operating earnings after 1998 can serve as a reasonable lower bound for valuation.  It’s unlikely that the market is cheaper than indicated by that measure, especially considering that the 10 year average earnings off of which the Total Return EPS CAPE is built incorporate the substantially elevated levels of corporate profitability associated with the 2005 to 2015 period.

Now, to the charts.  The following chart shows the hypothetical historical rolling average 10 year real total return that shareholders would have earned from January 1871 to March 2015 if valuations had stayed constant at Shiller CAPE levels of 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, and 30 respectively, with all dividends reinvested at those respective valuations.

histreturntr

Let’s start with the left portion of the chart.  If the Shiller CAPE had been equal to 5 throughout the entirety of U.S. market history, the average rolling 10 year real total return would have been north of 15%, more than double the actually realized value of roughly 6.3%.   Moving to the right, if the Shiller CAPE had been equal to 30 throughout the entirety of U.S. market history, the average rolling 10 year real total return would have been under 4%.  That’s with all other fundamentals held constant, the only differention coming from the different valuations at which the dividends are reinvested.  It goes without saying that the difference–more than 1100 real bps of return per year–matters.

To home in on the current situation, if the Total Return EPS CAPE formed on operating earnings (the lower bound) is the more accurate measure of valuation, and if, going forward, the market remains unperturbed at the current value of 21.93, then we should expect a future total return of 4.72%.  If the Total Return EPS CAPE formed on GAAP earnings (the upper bound) is the more accurate measure of valuation, and if, going forward, the market remains unperturbed at the current value of 25.96, then we should expect a future total return of 4.27%.

So there’s your range: 4.27% to 4.72%.  If valuation bulls win both the valuation argument and the profit margin argument, then that is the approximate return they should expect.  If they lose the valuation argument–if valuations fall back to the historical norm–then the drag described here will not apply.  But a new drag will be inserted: the drag of falling valuation, which pulls down returns by pulling down prices.

The following chart shows the same information as the previous chart, but with the returns expressed as deviations from the historical average rolling 10 year total return of 6.32%.

histr3

As you can see, a permanent CAPE of 5 would have added almost 9% to the actual average historical return.  A permanent CAPE of 30 would have subtracted almost 2.5% from the actual average historical return.  A permanent CAPE equal to the present value, somewhere between 21.93 and 25.96, would have subtracted anywhere from 1.60% to 2.05%.  That range represents a quarter to a third of the actual historical return.  So if market valuations stay where they are, on this “permanently elevated plateau”, investors should prepare to have anywhere from a quarter to a third of their future returns lopped off.

Now, valuation bulls might try to console themselves here by pointing out that reinvested dividends don’t matter as much to returns as they did in the past, since dividend payout ratios are substantially lower than they used to be. Wrong.  The same dividends are still being paid out, they’re just being paid out in a different form: in the form of share buybacks and corporate M&A.  The buybacks and M&A activities are being conducted at the same elevated valuations that the dividends would have been reinvested at.  So the result is the same.

In fact, if anything, the depressing future impact of permanently elevated valuations is likely to be more severe in the current market than it would have been in past markets.  In past markets, corporations grew EPS by investing in organic business growth, which isn’t affected by the market’s valuation.  The current era, however, is characterized by a reduced emphasis on organic business growth, and an increased emphasis on EPS growth through share count reduction–what is perjoratively termed “capital recycling.”

Currently, almost 100% of S&P 500 EPS is being devoted to capital recycling–some combination of dividends, buybacks, and buyouts.  The growth that this recycling will produce will be entirely determined by the market’s valuation–nothing else can make a difference to it but that.  And so if these simulations in historical data are telling us that we should mark down our future return expectations by 25% to 33% of the historical norm, then we should probably mark them down by an even greater amount, because the underlying allocation practice through which they will be driven down–capital recycling that occurs in lieu of organic business growth–is significantly more prevalent now than it was in prior eras.

Posted in Uncategorized | Comments Off on Capital Recycling at Elevated Valuations: A Historical Simulation

A New-and-Improved Shiller CAPE: Solving the Dividend Payout Ratio Problem

A common criticism of Professor Robert Shiller’s famous CAPE measure of stock market valuation is that it fails to correct for the effects of secular changes in the dividend payout ratio.  Dividend payout ratios for U.S. companies are lower now than they used to be, with a greater share of U.S. corporate profit going to reinvestment.  For this reason, earnings per share (EPS) tends to grow faster than it did in prior eras.  But faster EPS growth pushes up the value of the Shiller CAPE, all else equal.  Distortions therefore emerge in the comparison between present values of the measure and past values.

To give credit where it’s due, the first people to point out this effect–at least as far as I know–were Professor Jeremy Siegel of Wharton Business School and his former student, David Bianco of Deutsche Bank.  Siegel, in specific, wrote about the problem as far back as late 2008, during the depths of the financial crisis, when the Shiller CAPE was steering investors away from a market that he considered to be extremely cheap (see “Jeremy Siegel on Why Equities are Dirt Cheap”, November 18, 2008, link here).

In a piece from 2013, I attempted to demonstrate the effect with two tables, shown below:

shillerdiv2

shillerdiv1

The tables portray the 10 year earnings trajectories and Shiller CAPE ratios of two identical companies that generate identical profits and that sell at identical trailing-twelve-month (ttm) P/E valuations. The first company, shown in the first table, pays out 75% of its profit in dividends and reinvests the other 25% into growth (in this case, share buybacks that grow the EPS by shrinking the S). The second company, shown in the second table, pays out 25% of its profit in dividends, and reinvests the other 75% into growth.

As you can see, even though these companies are identically valued in all relevant respects, they end up with significantly different Shiller CAPEs.  The reason for the difference is that the second company reinvests a greater share of its earnings into growth than the first company.  Its earnings therefore grow faster.  Because its earnings grow faster, the act of “averaging” them over a trailing 10 year period reduces them by a greater relative amount.  Measured against that trailing 10 year average, the company’s price, appropriately set in reference to its ttm earnings, therefore ends up looking more expensive.  But, in truth, it’s not more expensive–its valuation is exactly the same as that of the first company.

The following chart illustrates the effect:

highlow

To summarize the relationship:

  • Lower Payout Ratio –> Higher Earnings Growth –> Higher CAPE, all else equal
  • Higher Payout Ratio –> Lower Earnings Growth –> Lower CAPE, all else equal

Now, how can we fix this problem?  A natural solution would be to reconstruct the CAPE on the basis of total return (which factors in dividends) rather than price (which does not). But that’s easier said than done.  How exactly does one build a CAPE ratio–or any P/E ratio–on the basis of total return?

Enter the Total Return EPS Index, explained here and here.  The Total Return EPS Index is a modified version of a normal EPS index that tells us, hypothetically, what EPS would have been, now and at all times in history, if the dividends that were paid out to shareholders had not been paid out, and had instead been diverted into share buybacks. Put differently, Total Return EPS tells us what earnings would have been if the dividend payout ratio had been 0% at all times.  In this way, it reduces all earnings data across all periods of history to the same common basis, allowing for accurate comparisons between any two points in time.

Crucially, in constructing the Total Return EPS, we assume that the buybacks are conducted at fair value prices, prices that correspond to the same valuation in all periods (equal to the historical average), rather than at market prices, which are erratic and often groundless.  To those readers who continue to e-mail in, expressing frustration with this assumption–don’t worry, you’re about see why it’s important.

The following chart shows the Total Return EPS alongside the Regular EPS from 1871 to 2015.  In this chart and in all charts presented hereafter, the index is the S&P 500 (and its pre-1957 ancestry), the values are appropriately inflation-adjusted to February 2015 dollars, and no corrections are made for the effects of questionable accounting writedowns associated with the last two economic downturns:

trp

Now, if all S&P 500 dividends had been diverted into share buybacks, then the price of the index would have increased accordingly. We therefore need a Total Return Price index–an index that shows what prices would have been on the “dividends become buybacks” assumption.

Calculating a Total Return Price index is straightforward.  We simply assume that the market would have applied the same P/E ratio to the Total Return EPS that it applied to the Regular EPS (and why would it have applied a different P/E ratio?). Multiplying each monthly Total Return EPS number by the market’s ttm P/E multiple in that month, we get the Total Return Price index.

In the chart below, we show the Total Return Price index for the S&P 500 alongside the Regular Price, from 1871 to 2015:

trp3

Generating a CAPE from these measures is similarly straightforward.  We divide the Total Return Price by the trailing 10 year average of the Total Return EPS.  The result: The Total Return EPS CAPE.

Shiller himself proposed a different method for calculating a CAPE based on total return in a June 2014 paper entitled “Changing Times, Changing Valuations: A Historical Analysis of Sectors within the U.S. Stock Market: 1872 to 2013” (h/t James Montier). The instructions for the method are as follows: Use price and dividend information to build a Total Return Index. Then, scale up the earnings by a factor equal to the ratio between the Total Return Index and the Price Index.  Then, divide the Total Return Index by the trailing ten year average of the scaled-up earnings.  In a piece from August of last year, I tried to build a CAPE based on Total Return using yet another method (one that involves growing share counts), and arrived at a result identical to Shiller.  The technique and charts associated with that method are presented here.

It turns out that both of these methods produce results identical to the Total Return EPS CAPE method, with one small adjustment: that we conduct the buybacks that form the Total Return EPS at market prices, rather than at fair value prices as initially stipulated. The following chart shows the three types of Total Return CAPEs together.  As you can see, the lines overlap perfectly.

fjksake

The three different versions of the CAPE overlap because they are ultimately doing the same thing mathematically, though in different ways.  Given that they are identical to each other, I’m going to focus only on the Total Return EPS version from here forward.  I’m going to refer to the version that conducts buybacks at fair value prices as “Total Return EPS (Fair Value) CAPE”, and the version that conducts buybacks at market prices as “Total Return EPS (Market) CAPE.”  I’m going to refer to Shiller’s original CAPE simply as “Shiller CAPE.”

The following chart shows the Total Return EPS (Market) CAPE alongside the Shiller CAPE, with the values of the former normalized so that the two CAPEs have the same historical average (allowing for a direct comparison between the numbers).

trmkt

(Note: in prior pieces, I had been comparing P/E ratios to their geometric means. This is suboptimal. The optimal mean for a P/E ratio time series is the harmonic mean, which is essentially what you get when you take an average of the earnings yields–the P/E ratios inverted–and then invert that average.  So, from here forward, in the context of P/E ratios, I will be using harmonic means only.) (h/t and #FF to @econompic, @naufalsanaullah, @GestaltU_BPG)

The current value of the Shiller CAPE is 27.5, which is 93% above its historical average (harmonic) of 14.2.  The current value of the Total Return EPS (Market) CAPE is 30.3, which is 71% above its historical average (harmonic) of 17.8.  Normalized to matching historical averages, the current value of the Total Return EPS (Market) CAPE comes out to 24.2.

At current S&P 500 levels, then, we end up with 27.5 for the Shiller CAPE, and 24.2 for the Total Return EPS (Market) CAPE, each relative to a historical average of 14.19. Evidently, the difference between the two types of CAPEs is significant, worth 12%, or 250 current S&P points.

But there’s a mistake in this construction.  To find it, let’s take a closer look at the chart:

mvout

From the early 1990s onward, the Total Return EPS (Market) CAPE (the red line) is significantly below the Shiller CAPE (the blue line), suggesting that the Shiller CAPE is overstating the market’s expensiveness, and that the Total Return EPS (Market) CAPE is correcting the overstatement by pulling the metric back down.

What is driving the Shiller CAPE’s apparent overstatement of the market’s expensiveness? The obvious answer would seem to be the historically low dividend payout ratio in place from the early 1990s onward.  All else equal, low dividend payout ratios push the Shiller CAPE up, via the increased growth effect described earlier.

But look closely.  Whenever the market is expensive for an extended period of time, the subsequent Total Return EPS (Market) CAPE (the red line) ends up lower than the Shiller CAPE (the blue line), by an amount seemingly proportionate to the degree and duration of the expensiveness.  Note that this is true even in periods when the dividend payout ratio was high, e.g, the periods circled in black: the early 1900s, the late 1920s, and the late 1960s.  If the dividend payout ratio were the true explanation for the deviations between the Shiller CAPE and the Total Return EPS (Market) CAPE, then we would not get that result.  We would get the opposite result: the high dividend payout ratio seen during the periods would depress the the Shiller CAPE relative to the more accurate total measures; it would not push the Shiller CAPE up, as seems to be happening.

The converse is also true.  Whenever the market is cheap for an extended period of time, the subsequent Total Return EPS (Market) CAPE (the red line) ends up higher than the subsequent Shiller CAPE (the blue line), by an amount seemingly proportionate to the degree and duration of the cheapness.  We see this, for example, in the periods circled in green: the early 1920s and the early 1930s through the end of the 1940s.  The deviation between the two measures is spatially small in those periods, but that’s only because the numbers themselves are small–single digits.  On a percentage basis, the deviation is sizeable.

The following chart clarifies:

hsit

So what’s actually happening here?  Answer: valuationnot the dividend payout ratio–is driving the deviation.  In periods where the market was cheap in the 10 years preceding the calculation, the Total Return EPS (Market) CAPE comes out above the Shiller CAPE. In periods where the market was expensive in the 10 years preceding the calculation, the the Total Return EPS (Market) CAPE comes out below the Shiller CAPE.  The degree above or below ends up being a function of how cheap or expensive the market was, on average.

The following chart conclusively demonstrates this relationship:

pinkgreen

The bright green line is the difference between the Total Return EPS (Market) CAPE and the Shiller CAPE as a percentage of the Shiller CAPE.  When the bright green line is positive, it means that the red line in the previous chart was higher than the blue line; when negative, vice-versa.  The pink line is a measure of how cheap or expensive the market was over the preceding 10 years, on average and relative to the historical average. When the pink line is positive, it means that the market was cheap; when negative, expensive.  The two lines track each other almost perfectly, indicating that the valuation in the preceding years–and not the payout ratio–is driving the deviation between the two measures.

What is causing this weird effect?  You already know.  The share buybacks associated with the Total Return EPS (Market) CAPE are being conducted at market prices, rather than at fair value prices.  The same is true for the dividend reinvestments associated with Shiller’s proposed Total Return CAPE and with the version I presented in August of last year; those reinvestments are being conducted at market prices.  That’s wrong.

When share buybacks (or dividend reinvestments) are conducted at market prices, then periods of prior expensiveness produce lower Total Return EPS growth (because the dividend money is invested at unattractive valuations that offer low implied returns).  And, mathematically, what does low growth do to a CAPE, all else equal?  Pull it down.  Past periods of market expensiveness therefore pull the Total Return EPS (Market) CAPE down below the Shiller CAPE, as observed.

Conversely, periods of prior cheapness produce higher Total Return EPS growth (because the dividend money is reinvested at attractive valuations that offer high implied returns).  And what does high growth do to a CAPE, all else equal?  Push it up.  Past periods of market cheapness therefore push the Total Return EPS (Market) CAPE up above the Shiller CAPE, as observed.

Looking at the period from the early 1990s onward, we assumed that the problem was with the Shiller CAPE (the blue line), that the low dividend payout ratio during the period was pushing it up, causing it to overstate the market’s expensiveness.  But, in fact, the problem was with our Total Return EPS (Market) CAPE (the red line).  The very high valuation in the post-1990s period is depressing Total Return EPS (Market) growth (the expensiveness of the share buybacks and dividend reinvestments shrinks their contribution), pulling down on the Total Return EPS (Market) CAPE, and causing it to understate the market’s  expensiveness.

The elimination of this distortion is yet another reason why the buybacks and dividend reinvestments that form the Total Return EPS (or any Total Return Index used in valuation measurements) have to be conducted at fair value prices, rather than at market prices.  Conducting the buybacks and dividend reinvestments at fair value prices ensures that they provide the same accretion to the index across all periods of history, rather than highly variable accretion that inconsistently pushes up or down on the measure.

Now, a number of readers have written in expressing disagreement with this point.  To them, I would ask a simple question: does it matter to the current market’s valuation what the market’s valuation happened to be in the distant past?

Suppose, for example, that in 2009, investors had become absolutely paralyzed with fear, and had sold the market’s valuation down to a CAPE of 1–an S&P level of, say, 50. Suppose further that the earnings and the underlying fundamentals had remained unchanged, and that investors had exacted the pummeling for reasons that were entirely irrational. Suppose finally that investors kept the market at the depressed 1 CAPE for two years, and that they then regained their senses, pushing the market back up to where it is today, in a glorious rally.  In the presence of these hypothetical changes to the past, what would happen to the current value of a Total Return EPS CAPE that reinvests at market prices?  Answer: it would go up wildly, dramatically, enormously, because the intervening dividends that form the Total Return index would have been invested at obscenely low valuations during the period, producing radically outsized total return growth.  What does high growth due to a CAPE? Push it up, so the CAPE would rise–by a large amount.

Is that a desirable result?  Do we want a measure whose current assessment of valuation is inextricably entangled in the market’s prior historical valuations, such that the measure would judge the valuations of two markets with identical fundamentals and identical prices to be significantly different, simply because one of them happened to have traded more cheaply or expensively in the past?  Obviously not.  That’s why we have to conduct the buybacks and reinvestments that make up the Total Return EPS at fair value.

The general rule is as follows.  When we’re using a Total Return index to model actual investor performance–what an individual who invested in the market would have earned, in reality, with the dividend reinvestment option checked off–we need to conduct the hypothetical reinvestments that make up the Total Return index at market prices.  But when we’re using a Total Return index to measure valuation–how a market’s price compares with its fundamentals–then we need to conduct the hypothetical reinvestments at fair value prices.

The following chart shows the Total Return EPS CAPE properly constructed on the assumption that the buybacks and reinvestments occur at fair value prices:

trepsfv

As you can see, the deviation between the two measures comes out to be much smaller. Normalized to the same historical average, the current value of the Total Return EPS (Fair Value) CAPE ends up being 25.9, versus 27.5 for the original Shiller CAPE.  The difference between the total return and the original measures comes out at 5.7%, a little over 100 current S&P points (versus 12% and 250 points earlier).

Surprisingly, then, properly reinvesting the dividends at the same valuation across history more than cuts the deviation in half, to the point where it can almost be ignored.  As far as the CAPE is concerned, when it comes to the kinds of changes that have occurred in the dividend payout ratio over the last 144 years, there appears to be little effect on the accuracy of Shiller’s original version.  The entire exercise was therefore unnecessary. Admittedly, this was not the result that I was anticipating, and certainly not the result that I was hoping to see.  But it is what it is.

It turns out that Shiller was right to reject the dividend payout ratio argument in his famous 2011 debate with Siegel and Bianco:

“Mr. Shiller did his own calculation about the impact of declining dividends on earnings growth and concluded that it is marginal at best, not meriting any adjustment.” — “Is the Market Overvalued?”, Wall Street Journal, April 9th, 2011.

If the subsequent foray into Total Return space caused him to change that view, then he should change it back.  He was right to begin with.  His critics on that point, myself included, were the ones that were wrong.

Now, this is not to suggest that we shouldn’t prefer to use the Total Return version of the CAPE over Shiller’s original version.  We should always prefer to make our analyses as accurate as possible, and the Total Return version of the CAPE is unquestionably the more accurate version.  Moreover, even though the changes in the dividend payout ratio seen in the U.S. equity space over the last 144 years have not been large enough to significantly impact the accuracy of the original version of the CAPE, the differences between the payout ratios of different countries–India and Austria, to use an extreme example–might still be large enough to make a meaningful difference.  Since the Shiller CAPE is the preferred method for accurately comparing different countries on a valuation basis, it only makes sense to shift to the more accurate Total Return version.  Fortunately, that version is simple and intuitive to build using Total Return EPS.

Admittedly, there is some circularity here.  In building the Total Return EPS Index on the assumption of fair value buybacks, we used the Shiller CAPE as the basis for estimating fair value.  If the Shiller CAPE is inaccurate as a measure of fair value, then our Total Return EPS index will be inaccurate, and therefore our Total Return CAPE, which is built on that index, will be inaccurate.  Fortunately, in this case, there’s no problem (otherwise I wouldn’t have done it this way). When you run the numbers, you find that the choice of valuation measure makes little difference to the final product, as long as a roughly consistent measure is used.  You can build the Total Return EPS Index using whatever roughly consistent measure you want–the Total Return CAPE will not come back appreciably different from Shiller’s original. What drove the deviations in the earlier charts were not small differences in the valuations at which dividends were reinvested, but large differences–for example, the difference associated with reinvesting dividends at market prices from 1942 to 1952, and then from 1997 to 2007, at prices corresponding to three times the valuation.

Now, there are other ways of adjusting for the impact of changing dividend payout ratios. Bianco, for example, has a specific technique for modifying past EPS values. As he explains:

“The Bianco PE is based on equity time value adjusted (ETVA) EPS.  We raise past period EPS by a nominal cost of equity estimate less the dividend yield for that period.”

I cannot speak confidently to the accuracy of Bianco’s technique because I do not have access to its details.  But if the method produces a result substantially different from the Total Return EPS CAPE (which it appears to do), then I would think that it would have to be wrong.  When it comes to changing dividend payout ratios, the Total Return EPS CAPE is airtight.  It treats all periods of history absolutely equally in all conceivable respects, perfectly reducing them to a common basis of 0% (payout).  Because it reinvests the dividends at fair value (the historical average valuation), every reinvested dividend in every period accretes at roughly the same rate, which corresponds to the actual average rate at which the market has historically accreted gross of dividends (approximately 6% real).

If our new-and-improved version of the CAPE is appropriately correcting the dividend payout ratio distortions contained in the original version, then the deviation between our new-and-improved version and the original version should be a clean function of that ratio (rather than a function of other irrelevant factors, such as past valuation).  When the dividend payout ratio is low, our new-and-improved version should end up below the original version, given that the original version will have overstated the valuation.  When the dividend payout ratio is high, our new-and-improved version should end up above the original version, given that the original version will have understated the valuation.

Lo and behold, when we chart the deviation between the two versions of the CAPE alongside the dividend payout ratio, that is exactly what we see: a near-perfect correlation (91%), across the full 134 year historical period.

delta

The blue line shows the difference between our new-and-improved version of the CAPE and the original version.  The red line shows the trailing Shiller dividend payout ratio, which is the 10 year average of real dividends per share (DPS) divided by the 10 year average of real EPS.  We use a Shillerized version of the dividend payout ratio to remove noise associated with recessions–especially the most recent one, where earnings temporarily plunged almost to zero, causing the payout ratio to temporarily spike to a value north of 300%.

The fact that the two lines overlap almost perfectly indicates that the deviation between our new-and-improved version and the original version is a function of the factor–the dividend payout ratio–that is causing the inaccuracy in the original version, rather than some other questionable factor.  That is exactly what we want to see.  It is proof positive that our new-and-improved version is correcting the distortion in question, and not introducing or exploiting other distortions (that, conveniently, would make the current market look cheaper).

Now, to be clear, the secular decline in the dividend payout ratio seen across the span of U.S. market history has not substantially affected the accuracy of the original Shiller CAPE.  However, it has substantially affected the trend growth rate of EPS.  So, though it may not be imperative that we use the Total Return version of the CAPE when measuring valuation, it is absolutely imperative that we use the Total Return version of EPS when analyzing earnings trends and projecting out future earnings growth.

We are left with the question: if the distortions associated with the dividend payout ratio are not significant, then why does the Shiller CAPE show the U.S. equity market to be so expensive relative to history?  We can point to three explanations.

  • First, on its face, the market just is historically expensive–even on a non-Shiller P/E measurement.  Using reported EPS, the simple trailing twelve month P/E ratio is roughly 20.5, which is 53% above its historical average (harmonic) of 13.4.  Using S&P corporation’s publication of operating EPS, the simple trailing twelve month P/E ratio is 18.8, which is 40% above that average.
  • Second, the accounting writedowns associated with the 2008-2009 recession are artificially weighing down the trailing average 10 year EPS number off of which the Shiller CAPE is calculated.  Prior to 2014, this effect was more significant than it is at present, given that the 2001-2003 recession also saw significant accounting writedowns.  The trailing 10 year average for the years up to 2014 therefore got hit with a double-whammy.  That’s why the the increase in the Shiller CAPE in recent years has not been as significant as the increase in market prices (since December 2012, the CAPE is up roughly 30%, but prices are up roughly 50%).  2014 saw the 2001-2003 recession fully drop out of the average, reducing the CAPE’s prior overstatement.
  • Third, as the chart below shows, real EPS growth over the last two decades–on both a regular and a Total Return basis–has been meaningfully above the respective historical averages, driven by substantial expansion in profit margins.  Recall that high growth produces a high CAPE, all else equal.

profmarginincluded

These last two factors–the effects of accounting writedowns and the effects of profit margin expansion–will gradually drop out of the Shiller CAPE (unless you expect another 2008-type recession with commensurate writedowns, or continued profit margin expansion, from these record levels).  As they drop out, the valuation signal coming from the Shiller CAPE will converge with the signal given by the simple ttm P/E ratio–a convergence that is already happening.

We conclude with the question that all of this exists to answer: Is the market expensive? Yes, and returns are likely to be below the historical average, pulled down by a number of different mechanisms.  Should the market be expensive?  “Should” is not an appropriate word to use in markets.  What matters is that there are secularsustainable forces behind the market’s expensiveness–to name a few: low real interest rates, a lack of alternative investment opportunities (TINA), aggressive policymaker support, and improved market efficiency yielding a reduced equity risk premium (difference between equity returns and fixed income returns).  Unlike in prior eras of history, the secret of “stocks for the long run” is now well known–thoroughly studied by academics all over the world, and seared into the brain of every investor that sets foot on Wall Street.  For this reason, absent extreme levels of cyclically-induced fear, investors simply aren’t going to foolishly sell equities at bargain prices when there’s nowhere else to go–as they did, for example, in the 1940s and 1950s, when they had limited history and limited studied knowledge on which to rely.

As for the future, the interest-rate-related forces that are pushing up on valuations will get pulled out from under the market if and when inflationary pressures tie the Fed’s hands–i.e., force the Fed to impose a higher real interest rate on the economy.  For all we know, that may never happen.  Similarly, on a cyclically-adjusted basis, the equity risk premium may never again return to what it was in prior periods, as secrets cannot be taken back.

Posted in Uncategorized | Comments Off on A New-and-Improved Shiller CAPE: Solving the Dividend Payout Ratio Problem

Using Total Return EPS to Decompose Historical S&P 500 Performance: Charts from 1871 to 2015

10d

In this piece, I’m going to do five things:

  • First, I’m going to clarify the purpose of Total Return EPS, what it’s trying to accomplish.  In a single sentence, the purpose of Total Return EPS is to convert dividends into EPS so that the fundamental sources of return can be added together into one single term whose past growth rate can be analyzed and whose likely future growth rate can be projected.
  • Second, I’m going to explain why the trend growth rate of Total Return EPS for the S&P 500 (~6%) is roughly equal to the historical average return on equity for the U.S. corporate sector (~6%).  The explanation will include a proposed theory for why return on equity generally reverts to the mean, and also for why it may not revert to its prior mean in the present environment.
  • Third, I’m going to address a question that a significant number of readers have asked: why does Total Return EPS assume that buybacks occur at fair value, rather than at market prices?
  • Fourth, I’m going to show how actual total return can be “decomposed”–i.e., separated out–into three contributing components: (1) Total Return EPS growth, which consists of regular EPS growth plus the return from reinvested dividends (or hypothetical share buybacks), (2) the return contribution from the change in valuation–in this case, the change that occurs in the ttm P/E ratio from price to sale, and (3) the return contribution from interim deviations from fair value–a neglected source of return that arises from the valuations at which dividends are reinvested (or at which shares are hypothetically repurchased), and therefore the rate at which they accrete.
  • Fifth, I’m going to present charts of these components for the S&P 500 from 1871 to 2015, on time horizons of 10, 20, 30, 40, 50, 60, and 70 years.

Three Options: Dividends, Expansion, Repurchases

When the corporate sector earns profit, it can do one of two things: distribute the profit to shareholders as dividends, or reinvest the profit.

  • When it distributes the profit to shareholders as dividends, the shareholders get a direct return–a direct deposit of money into their accounts.
  • When it reinvests the profit, the shareholders get an indirect return–“growth.” The profits earned in future periods, and the future dividends that can be paid from them, increase in size.  In an efficient market, this increase coincides with an increase in the market prices of shares, allowing shareholders to realize a return by selling.

Looking closer at the second option, the corporate sector can reinvest profit in one of two ways: by using it to fund business expansion, or by using it to repurchase equity (or debt). Both options produce growth in earnings per share (EPS).

  • When the corporate sector uses profit to fund business expansion, it adds new capital that it can use to produce and sell additional output to the economy, from which additional income can be earned.  It grows the EPS by growing the E.
  • When the corporate sector uses profit to repurchase equity–for example, by buying back shares on the open market and then cancelling them–it grows the EPS by shrinking the S.  (Note: The corporate sector can also use profit to repurchase or retire debt.  We can view this option as roughly equivalent to the repurchase of equity. Both options entail a reduction in the number of outstanding claims on the corporate sector, rather than an increase in the size of the corporate sector’s operations).

What we have, then, are three destinations for corporate profit: (1) the payment of dividends, (2) investment in business expansion, and (3) the repurchase of equity (or debt).  The first option entails a direct return, a direct deposit of money into shareholder pockets.  The second and third options entail an indirect return, achieved through growth in EPS.

EPS Growth: In Search of a Trend

What we want to know is the “trend” (or “normal”) rate of growth of EPS.  Knowing that trend rate would allow us to roughly estimate the likely future trajectory of EPS, given its position relative to trend.

To illustrate, suppose that the trend rate of EPS growth is 4% per year, but that EPS over the last several years hasn’t grown at all, or worse, has fallen substantially.  We would then expect future EPS growth to be higher than the trend rate, higher than 4%, as EPS catches up.  We would expect there to have been some kind of stunting process–say, a depression in profit margins–that explains the underperformance relative to trend, and that entails the potential for future outperformance, to be unleashed in an eventual mean-reversion.

The problem, of course, is that when we look at the historical data, we do not find a stable, reliable trend growth rate in EPS.  Instead, we find a trend growth rate that has increased substantially over time.  The following table shows averages of rolling 10 year annualized real EPS growth rates for the S&P 500 for the periods 1871 to 1930, 1930 to 1990, and 1990 to 2015, with each period beginning and ending in January:

epsperiods

As  you can see in the table, the average rolling growth rate seen from 1990 to 2015 is four times the rate seen 100 years before it.  And note that this rate is the growth rate of GAAP EPS.  It include the effects of the questionable accounting writedowns that took place in 2003 and 2009.  If we use a corrected version of EPS that excludes those writedowns, the rolling average growth rate for the period increases to 4.76%–more than six times the rate achieved 100 years before it.

The reason for the increase in the trend growth rate of EPS is no mystery.  EPS growth varies inversely with the share of profit that is paid out as dividends.  That share has fallen over time.  The following table shows average payout ratios for the periods in question:

avgporat

Now, there’s a legitimate question to ask here.  How much EPS growth should a given reduction in the dividend payout ratio actually produce? Is the observed increase in the growth rate, from 0.72% to 3.16%, commensurate with the observed decrease in the average payout ratio, from 69% to 52%?  Sadly, there’s no way to know.  Because we have no way to know, we can’t be sure that the reduced payout ratio is the only factor, or even the main factor, driving the increased growth.  For all we know, there could be some other factor driving it, a factor that will substantially impact growth going forward, either positively or negatively.

Return on Equity and Fundamental Total Return

We can distinguish between two sources of shareholder total return.  The first source is “fundamental”, and arises from the payment of dividends and the growth of EPS (or some other relevant fundamental).  The second source is “nonfundamental”, and arises from changes that occur in the valuations of assets between the time of purchase and the time of sale.  If you buy an asset and it pays you a dividend, or its price goes up in response to growth in relevant fundamentals–sales, earnings, net asset values, and so on–that is a “fundamental” return.  If you buy an asset and its price goes up independently of any type of growth in fundamentals–somebody just offers to pay you a higher price for the asset, because they want it more than the next person–that is a “nonfundamental” return.

The item that follows a consistent trend over time is not EPS growth per se, but the fundamental total return that accrues to shareholders–the return that dividends and EPS growth combine to produce.  Let me now explain why that return follows a consistent trend.  Bear with me.

The fundamental total return that accrues to shareholders is a function of the return that corporations generate on their equity, on the amount of capital that was invested to form them.  That return, after all, has to go to someone; it goes to the shareholders, those that made the investment, that put the capital in.

Now, return on equity (ROE) is mean-reverting.  When ROE is high in a given sector or industry, new investment flocks in, seeking to capture the high return.  The new investment leads to excess capacity, increased competition, weakened pricing power, and a reduction in profit that pulls the ROE for the sector or industry back down. When ROE is low in a given sector or industry, new investment stops happening.  The reduction in investment leads to an eventual undercapacity, reduced competition, increased pricing power for the remaining firms, and an increase in profit that pushes the ROE for the sector or industry back up–assuming, of course, that the goods and services being produced are actually wanted by the economy.  If they are not wanted, then the ROE for the industry or sector will go to zero, which is where it belongs for those that make unwanted things.

We’re currently seeing a textbook case of this process play out in the energy sector.  The economy needs a certain amount of oil.  The prior market price of oil–$75+–reflected the marginal cost of producing that amount, plus the extra “oomph” that speculation probably added.  But then efficient new drilling techniques were developed.  The strong profits that these techniques could earn with oil at $75+ led to an investment boom.  The investment boom eventually created an overcapacity that has pushed the price of oil down and that has dramatically lowered the return.  New investment has therefore dried up–and will stay dried up–until an undercapacity develops that increases the price enough to make the return attractive again.

This process of mean-reversion functions at its fiercest in the energy sector, where the good being sold is a pure commodity, and where there are few barriers or “moats” to block out competition and new entry.  But it applies in a general sense to all sectors and industries, and to the aggregate corporate sector as well.

(Caveat: If an economy evolves in a way that entails an increase in the number of barriers and “moats” in place to block out competition and new entry–i.e., in a way that makes it harder for new capital to partake in the high returns that existing capital might be enjoying–then the “mean” that the return on equity reverts to might increase accordingly. It remains an open question as to whether the new technology economy, with its tendency to produce winner-take-all scenarios in which the first mover is forever protected from competition–think $MSFT, $FB, $GOOG, $AAPL, and so on–has provoked such an increase.  I suspect that there is at least some of that effect at play in the much-discussed increase in ROEs and profit margins that we’ve seen take place over the last 20 years.)

Now, because the fundamental total return to shareholders is a function of the ROE, and because the ROE is mean-reverting, the fundamental total return to shareholders–paid out to them in dividends and growth–is similarly mean-reverting. Its mean-reverting nature is the reason that it follows a reliable trend over time.

Some might find this point hard to grasp–it’s admittedly hard to explain. To get a better feel for it, just think about the fundamental return that accrues to energy sector shareholders–shareholders in companies like $XOM and $CVX.  Can you see how the process that produces mean-reversion in energy sector ROEs would also produce mean-reversion in the fundamental return that $XOM and $CVX shareholders receive over time?  The same operating environment that allowed those companies to generate outsized fundamental returns–outsized EPS growth and outsized dividends–when oil was $75+ is what pulled in all of the new investment that fueled the current overcapacity, the squeeze to find buyers of all of the output, that is now pushing those same fundamental returns back down, hedges notwithstanding.

If there had been no new oil to drill, then that would have been a very powerful “moat”, and the high returns that these companies enjoyed might have been sustainable.  But when there is a new discovery that opens up ample new supply with the promise of a high return to anyone with capital that wants to make an investment in it, the high returns–to the new entrants and the existing players–simply will not last.

Ideally, then, we would ditch the effort to find the trend growth rate in EPS, and would instead focus on finding the trend in the metric that actually follows a trend–fundamental total return.  The problem, of course, is that fundamental total return is tied up in two distinct types of terms: EPS growth and dividends.  To properly analyze the trend in that return over time, we need a way to convert the terms into the same type of term, so that they can be added together to produce a single term, a single index.

The optimal way to solve the problem is to convert the dividends into a type of additional EPS, and then add the additional EPS to the actual EPS.  Then, we will end up with one single term that grows over time at a consistent trend rate, whose position relative to trend we can examine and make informed future projections based on.

That is precisely what the technique in the prior piece tries to do.  It tries to convert dividends into additional EPS by hypothetically assuming that dividends are diverted into share buybacks.  It then adds the additional EPS from the hypothetical share buybacks to the EPS that actually occurred, so as to form the unified, all-in-one term being sought: Total Return EPS.

The Equivalence of Reinvested Dividends and Share Buybacks

Now, from a total return perspective, it doesn’t matter whether a corporation chooses to distribute its profit as dividends, or use its profit to buy back shares.

  • If it pays out dividends, the dividends will be reinvested (that’s at least the assumption that “total return” indices hypothetically make).  The reinvestments will cause the number of shares that each shareholder owns to grow.
  • If it buys back shares, its outstanding share count will shrink, and therefore its earnings per share (EPS) will grow.  Mathematically, the growth in its EPS will roughly equal the growth in the number of shares that the shareholder would have come to own via dividend reinvestment.  If the market is efficient–meaning that it properly prices value–then the shareholder will end up no better or worse off, at least on a pre-tax basis (after tax, of course, is a different story).

Another way to express the point: When a corporation buys back shares with money that would otherwise have gone to dividends, it is effectively doing the dividend reinvestments for the shareholders.  It is accumulating shares in their names, as opposed to paying money out to them for them to accumulate shares on their own, independently of the company.  In truth, the two are not perfectly equivalent–share buybacks are actually slightly more accretive than reinvested dividends, for mathematical reasons that are too tedious to try to explain.  But they are close enough.

For any equity market, then, we are free to interchange dividends and share buybacks at will.  We can rebuild total return indices on the assumption that all dividends are hypothetically replaced with share buybacks, or that all share buybacks are hypothetically replaced with dividends–the replacements, if properly constructed, will have no perceptible effect on the total return.

The technique used to build the Total Return EPS index exploits this convenient equivalence.  It assumes, hypothetically, that for all of history, all dividend money that actually got paid was not actually paid, and was instead used to buy back shares.

When we examine Total Return EPS over history, we find that it does follow a reliable trend, as theory would suggest.  Consider the following table, which shows average rolling 10 year EPS and Total Return EPS growth rates over the periods identified in the previous table.

avg

With Total Return EPS, we see a far more consistent growth rate.  That growth rate is roughly on par with the corporate sector’s average historical return on equity, some number close to 6%, as theory would again suggest (FRED).

6percent

Buybacks: Why Fair Value Prices?

The hypothetical buybacks that are used to form the Total Return EPS are assumed to occur not at market prices, but at fair value prices–prices that correspond to an average valuation across history.  A number of readers have tweeted and e-mailed in, asking why we make this assumption.  Why not assume that the buybacks occur at market prices instead, and save the confusion?

The reason is simple.  We’re trying to build an index that captures the fundamental total return that corporations generate for their shareholders through the profits they earn, which they deliver to their shareholders in the form of dividends and EPS growth.  If we were to conduct the buybacks at market prices, then that return would fluctuate based on the market’s valuation–a nonfundamental factor that has nothing to do with those profits.

In conducting the hypothetical buybacks, we are effectively converting dividends into a type of EPS (that gets added to the regular EPS to form the Total Return EPS).  That is, instead of paying out the dividends, we are using them to shrink the S, which effectively adds more EPS (makes EPS bigger by reducing the denominator).

Now, the valuations at which the buybacks occur represents the effective rate of conversion between dividends and EPS.  A low buyback valuation will convert dividends into a large amount of EPS, as the dividends will buy back a large number of shares. Conversely, a high buyback valuation will convert dividends into a small amount of additional EPS, as the dividends will buy back only a small number of shares.

To capture the true fundamental total return, and not add nonfundamental noise associated with where market prices just so happen to be, we need to ensure that the same rate of conversion between dividends and EPS is applied in all periods.  We need to ensure that each dividend, adjusted for size, adds the same amount of relative EPS as every other, regardless of when it happens to be paid.  That’s why we assume that all buybacks occur at the same valuation: “fair value”, which in the previous piece, we defined in terms of the historical average of the Shiller CAPE).

The assumption that the buybacks that underlie Total Return EPS occur at fair value may seem trivial and unimportant, but it makes a meaningful difference.  This difference will become particularly significant when we try to use Total Return EPS to build a new-and-improved Shiller CAPE.  If we use market valuations for the buybacks in Total Return EPS, the new-and-improved Shiller CAPE will end up skewed, giving an unnecessarily inaccurate picture of the market’s true valuation.  I plan to discuss the point in more detail in a later piece on the Shiller CAPE.

An Important Clarification: The Buybacks are Hypothetical

Let me now make an important clarification, based on some of the questions I’ve received. Regular S&P 500 EPS has continually grown throughout history.  Its growth has been driven by both business expansion (real economic investment that adds capital and increases output–EPS growth driven by growth in the E) and the repurchase of equity (buybacks, acquisitions, mergers, and so on–EPS growth achieved by shrinkage of the S).

The Total Return EPS doesn’t modify any of that growth.  Rather, what Total Return EPS does is add the additional growth that would have been produced if the dividends that were paid out had instead been used, completely hypothetically, to buy back shares (or fund acquisitions, mergers, and so on).

The following schematic makes the point more clear:

chart

Some readers have asked: in constructing Total Return EPS, why do you assume that the buybacks occur at fair value prices, when, in reality, corporations like Apple and IBM are buying back their shares at market prices?   This question misses the point.  When I talk about conducting buybacks at fair value prices, I’m not referring to those buybacks, the ones that actually happened in reality, or that are happening now.  Their effects have already shown up, or will show up, in regular EPS growth.  The buybacks that I’m referring to, the ones associated with the construction of Total Return EPS, are hypothetical buybacks–buybacks that didn’t actually happen, that aren’t happening, but that we assume happened or are happening, in lieu of dividends, so as to convert the dividend return into a type of EPS growth.

Interim Valuation: A Neglected Driver of Returns

Our assumption that the hypothetical buybacks occur at fair value highlights a crucial fact about returns that often gets missed.  Valuations matter to returns not only in relation to terminal prices–the price at which you buy and the price at which you sell–but also in relation to interim prices–the prices at which your dividends get reinvested (or, in this context, at which your CEO buys back shares in your name).  As we will later see, this effect is not small, not negligible, even though we might intuitively expect it to be.

In a future piece, I’m going to explore the impact further.  For a quick teaser, consider the following surprising result.  From 1871 to 2015, the actual annualized Total Return for the S&P 500–including the return from changes in valuation–was 6.89%. If, from 1871 to 2015, everything had been kept the same, except that interim prices had been permanently pushed up to a Shiller CAPE equal to the current value of 27.5, with the dividends reinvested at those high prices, rather than at the much cheaper prices that were actually realized, the total return would have been only 4.78%.  That’s more than 200 bps–almost a third of the historical total return–lost to this mechanism.

Remember this fact the next time you find yourself assuming that a policymaker-coddled market that always stays elevated, that never crashes or corrects, would somehow be a good thing for buy-and-hold investors.  It would not be.  The winners in such a market would actually be the impatient, weak-willed, market-timing-prone people who sell to buy-and-hold investors, when those investors go to reinvest their dividends (or when corporations go to buy back shares, which is all they seem to want to do these days).  Those people would never again have to sell at unfair prices, never again have to foot the bill for the bargains that buy-and-hold investors–the Warren Buffets of the world–have historically enjoyed.

In the next section, I’m going to present the theory that underlies the decompositions that will follow at the end of the piece, so that others can reproduce the results themselves.  If you’re not interested, feel free to fast forward to the end, where the charts are presented and discussed.  To briefly summarize, I’m going to arrive at the following two equations:

(6) Total Return EPS Growth = EPS Growth + Return Contribution from Dividends Reinvested at Fair Value

(7) Total Return = Total Return EPS Growth + Return Contribution from Change in P/E Ratio + Return Contribution from Interim Deviations from Fair Value

Along the way, I’m going to explain what each term means, and how each term is calculated from the data.

Decomposing Equity Total Returns: The Theory

In his 1981 magnum opus, Robert Shiller eloquently delineated the fundamental components of equity total return:

“Once we know the terminal price and intervening dividends, we have specified all that investors care about.” — Robert Shiller, “Do Stock Prices Move Too Much to be Justified by Subsequent Changes in Dividends?”, 1981 

We can translate this point loosely as follows:

(1) Total Return = Price Growth + Return Contribution from Reinvested Dividends

We can express price growth in terms of growth in EPS (a fundamental that gets decided by economic processes) and the return contribution from the change in the P/E ratio (a value that gets decided based on the brute forces of supply and demand mixed together in equilibrium with investor beliefs about what is a fair, appropriate, justified, responsible, sufficiently-rewarding price to pay).

(2) Price Growth = EPS Growth + Return Contribution from Change in P/E Ratio

Substituting (2) into (1) we get:

(3) Total Return = EPS Growth + Return Contribution from Change in P/E Ratio + Return Contribution from Reinvested Dividends

Now, let’s look at this last term, Return Contribution from Reinvested Dividends.  We can express this term as the combination of (a) the Return Contribution from Dividends Reinvested at Fair Value and (b) the Return Contribution from Interim Deviations from Fair Value.  The return contribution from reinvested dividends is the return that would have accrued if they had been reinvested at fair value, plus the “extra” return (positive or negative) that has arisen from the fact that, in reality, they were not actually reinvested at fair value, but were reinvested at higher or lower valuations, producing a lower or higher return.

We end up with:

(4) Return Contribution from Reinvested Dividends = Return Contribution from Dividends Reinvested at Fair Value + Return Contribution from Interim Deviations from Fair Value

Combining (3) and (4) we get a total return equation with four components:

(5) Total Return = EPS Growth + Return Contribution from Dividends Reinvested at Fair Value + Return Contribution from Change in P/E Ratio + Return Contribution from Interim Deviations from Fair Value

Now, to substitute in Total Return EPS, we recall that Share Buybacks and Reinvested Dividends are the same thing.  This means:

(6) Total Return EPS Growth = EPS Growth + Return Contribution from Dividends Reinvested at Fair Value

Inserting (6) into (5) we get:

(7) Total Return = Total Return EPS Growth + Return Contribution from Change in P/E Ratio + Return Contribution from Interim Deviations from Fair Value

These two equations, (6) and (7), are the equations that we are going to visually plot. Before we can do that, however, we need to find a way to quantify the terms in each equation.

We do that as follows:

    • (Regular) EPS Growth: Trivial.  We calculate the annualized % change between the starting and finishing values of (regular) EPS.
    • Total Return EPS Growth: Again, trivial.  We calculate the annualized % change between the starting and finishing values of Total Return EPS.  Directions for how to build the Total Return EPS index can be found in the previous piece.
    • Return Contribution from Dividends Reinvested at Fair Value: We take the difference between Total Return EPS Growth and Regular EPS Growth.  This difference equals the contribution from reinvested dividends (or, alternatively, the contribution from share buybacks–they are the same thing).
    • Return Contribution from Change in P/E Ratio: We take the difference between price growth and EPS Growth.  This difference just is the return contribution from the change in the P/E ratio.

Now, to get the final term, the Return Contribution from Interim Deviations from Fair Value, we need to build a new index.  Call that index the “Total Return EPS with Purchases at Market Prices” index.  This index is identical to the Total Return EPS index, except that the buybacks are conducted (or the dividends reinvested) at market prices rather than at fair value prices.

    • Return Contribution from Interim Deviations from Fair Value: Take the difference between the annualized growth of “Total Return EPS with Purchases at Market Prices” and the annualized growth of Total Return EPS.  This difference just is the added return that comes from buying back shares (or reinvesting dividends) at market valuations that do not always average to fair value.

Charting the Decomposition

In the following charts, I’m going to decompose–i.e., separate out–the historical S&P 500 total return into three contributing components: Total Return EPS Growth (purple), Return Contribution from Change in P/E Ratio (orange), and Return Contribution from Interim Deviations from Fair Value (blue).  I’m going to further decompose Total Return EPS Growth into two components: Return from Reinvested Dividends (identical to Return from Hypothetical Share Buybacks) (green) and Regular EPS Growth (yellow). Recessionary periods for the U.S. economy will be shaded in gray.

The decompositions will be conducted on the returns at time horizons of 10, 20, 30, 40, 50, 60, and 70 years, from 1871 to 2015.  For each time horizon, there will be two separate charts (miniatures shown below), with the first chart decomposing the total return, and the second chart decomposing the Total Return EPS.  As with all numbers in this piece, the growth rates and returns are real, properly adjusted for inflation.

10 Years

10d

A brief discussion on how to read the chart.  The x-axis has two dates.  The upper is the starting date for a period, the lower is the ending date–in this case, 10 years later.

Consider the slice of the chart that begins with 1989 and ends with 1999.  I’ve boxed it in red below:

10dd

The purple, the Total Return EPS growth, was roughly on par with the historical average, around 6%. What this means is that from 1989 to 1999, the sum of the return from dividend reinvestments (or hypothetical share buybacks–same thing) at fair value and the return from regular EPS growth amounted to 6% per year.

The orange, the Contribution from the Change in P/E ratio, was enormously positive, adding more than 10% to the return.  Of course, that’s consistent with what we remember. In 1989, valuations were reasonable; in 1999, they were in a bubble.  The transition from normal valuations to bubble valuations produced phenomenal returns for shareholders. In hindsight, of course, nothing was actually “produced”–returns were simply pulled forward from the future, stolen from those that bought in at the end.

The blue, the Contribution from Interim Deviations from Fair Value, was actually negative, subtracting approximately 1% from the return.  This also checks with what we remember. From 1989 to 1999, valuations were substantially above average.  There were only a few very mild corrections that took place–certainly nothing resembling a crash.  For the most part, the market just went straight up.  The above average valuations depressed the return from reinvested dividends relative to the alternative of a market at fair value (which is what Total Return EPS is indexed to).

Notice that as we move to the right in the chart, towards starting dates in the early 1990s, the Contribution from Interim Deviations from Fair Value gets even more negative, approaching -2% per year.  To understand why, recall that the market in the late 1980s and early 1990s was actually valued fairly attractively.  When we move to the right, those years drop out, and get replaced by the acute phase of the tech bubble, when the market was radically expensive.

The thin black line is the actual total return, which almost exactly equals, within a few bps, the sum of the contributors, as it should.  Note that I’m calculating the actual total return not by summing the contributors, but by building an entirely separate total return index, using the normal methods for doing so.  The chart can therefore be taken as empirical proof that the decomposition is analytically correct–the numbers, calculated by separate methods, add up perfectly, as they should.

The chart of the decomposition of Total Return EPS, shown below, follows the same structure:

10c

The chief thing to notice in the Total Return EPS chart is how the mix of the return has shifted from green (reinvested dividends, or alternatively, hypothetical share buybacks) to yellow (EPS growth).  This shift will become more clear and compelling as we move to longer time horizons, where the interfering cyclical noise will get smoothed out.

20 Years

20d

20c

30 Years

30d

30c

40 Years

40d

40c

50 Years

50d

50c

60 Years

60d

60c

70 Years

70d

70c

Conclusions

On longer time horizons, we see certain patterns crystallize.  The Total Return EPS, shown in purple, converges on a trend growth rate slightly below 6% annualized.  The green–the dividend (buyback) return–shrinks, while the yellow–the return from regular EPS growth–expands, keeping the sum of the two–Total Return EPS growth–on trend.

The shift from green to yellow is the shift in corporate preference visualized–away from dividends and towards growth.  When we try to conduct trend analysis on regular EPS–the yellow–we inevitably miss this shift, and therefore arrive at faulty conclusions.  What we need to analyze instead is the black line, the sum, the Total Return EPS, which has held to its trend comparatively well over the long-term.

In earlier periods of the charts, frequent market cheapness contributed meaningfully to the return.  But over time, as the market has become more efficient, less prone to violent downturns and crashes, that contribution has faded.  Notice that the blue–the contribution from interim deviations from fair value–is much thinner now than it used to be.  In charts of shorter time horizons (10 or 20 years, for example), it has even gone negative.  The shift to a negative contribution reflects the secular increase that has occurred in the market’s valuation, the valuation at which dividends are reinvested.  If, going forward, the market successfully holds steady at its currently elevated valuation, successfully avoiding the pull of downturns and crashes, then the blue will stay negative, and total returns will underperform the historical average accordingly–simply by that mechanism, never mind the others.

Investors need to understand that they can’t have it both ways: they will have to either accept historical levels of volatility, which will allow them to reinvest their dividends at cheap prices every so often (and allow their CEOs to buy back shares and acquire companies at those same prices), or they will have to accept lower than normal historical returns.  The growing corporate preference for buybacks (and acquisitions, and mergers) as a low-risk, tax-efficient alternative to risky capital expenditure will only exacerbate this impact.

At present, nearly 100% of current S&P 500 EPS is being used to fund dividends and buybacks–a trend that looks set to continue.  Going forward, interim valuations–which will influence the returns that those dividends and buybacks produce–are therefore likely to be even more impactful than they were in the past.  If valuations remain where they currently are–at levels that would qualify as historically expensive even on the uncertain assumption that profitability will remain at record highs–future returns are likely to suffer accordingly.

Posted in Uncategorized | Comments Off on Using Total Return EPS to Decompose Historical S&P 500 Performance: Charts from 1871 to 2015

Introducing the Total Return EPS Index: A New Tool for Analyzing Fundamental Equity Market Trends

rshillerIn late December of 2010, with the S&P 500 pushing through the mid 1200s on the heels of QE2 exuberance, my favorite financial economist–the great Robert Shiller–made what will likely turn out to be a very inaccurate prediction.  To be fair to Shiller, it wasn’t really a “prediction”, but more of an “estimate.” He estimated that the S&P 500 would trade at 1430 in the year 2020.

Wait, did he mean 2430?  No.  He meant 1430–the level the market was at in late 2012, just before it went on its epic run.  He estimated that the market would be there eight years later, in 2020.

His estimate could still end up being right.  But, at this point, it will take a lot of luck. The trough of a 30% market downturn–from current levels, not higher levels–will have to occur exactly in 2020.  Either that, or the market will have to suffer an even larger downturn, and then recover to the target by 2020.  Neither of these chance occurrences seems to have been intended as the basis for the original estimate.

Rather than chuckle at the wrongness of the estimate, we may want to consider its implications for the current market.  Here we have an excellent financial economist, one of the few to have accurately identified both the tech bubble and the housing bubble in real-time, estimating that in five years the market will be 30% below where it is today.

What is that telling us?  Maybe it’s telling us that nobody–not even the “experts”–knows what’s going to happen in the market, and that all of this “analysis” mumbo-jumbo is just a front that people offer up in defense of stances that they’re already entrenched in for other reasons (emotional, dispositional, moral, ego-related, career-related, business-strategy-related, because they’ve already gone on record, cemented an identity, staked a reputation, and need to be “right”, etc.)

But maybe it’s telling us something else.  Maybe it’s telling us that this market has been stretched very thin, well beyond the limits of defensible valuation, such that well-reasoned prior estimates for where prices would now be are missing the mark by miles. If true, that’s great news for those that were buying equities in 2010.  It’s not great news for those that are buying equities now.

Historical EPS: The Trend is Not Your Friend

In the CNBC interview, Shiller explained the basis for his estimate:

“The problem with the traditional price earnings ratio is that earnings are just too volatile from year to year…  We’re talking ten years out.  So I’m going to go back to 100 years.  The growth of real inflation directed earnings is surprisingly low.  From 1890 to 1990 it was only 1.5 percent a year…  I take earnings and I extrapolate them out at 1.5 percent from where they–S&P 500 earnings–are now and then I apply a price earnings ratio of 15, which is the historical average for 1890 to 1990.”  — Robert Shiller, CNBC, December 31st, 2010 

The best way to understand Shiller’s logic here would be to refer to a chart of S&P 500 earnings per share (EPS) relative to its historical “trend.” The CEO of Business Insider, Henry Blodget, recently highlighted a nice version of such a chart, put together by GMO’s James Montier.

montierdaChris Brightman of Research Affiliates presented a similar chart in a 2014 piece entitled “The Profits Bubble.”  I’ve written over the chart in maroon:

spxshiller2

The logic works like this.  We assume that EPS oscillates around, and eventually reverts to, a long-term trend (the gold line in the first chart, the black line in the second).  To estimate EPS for a date in the future, we determine where the trend will be on that date (that is, we assume an eventual reversion to it).  We then apply a “normal” P/E multiple to the EPS estimate to get an estimated price.

As of 2Q 2010, reported S&P 500 EPS was $67.10, almost perfectly on the historical trend.  So Shiller extrapolated the trend growth rate–1.5% real, plus 2% inflation–out 10 years.  He then applied a 15 P/E ratio.  At 1.5% + 2% = 3.5% growth per year over 10 years, $67.10 becomes $95.  15 times $95 is roughly 1430–Shiller’s 2020 price target.

Before I share my chief concern with this logic, let me say that I agree with Shiller’s (and Blodget’s, and Montier’s, and Brightman’s) thesis that long-term returns from current prices will likely be disappointing, meaningfully less than the historical average of 6% real. I would much rather be with them in that debate than with those that are arriving at optimistic conclusions by naively extrapolating their own recent experiences, pretending that markets are always as kind and rewarding as they have been to U.S. investors over the last 6 years.  As a strategy, extrapolation can work, but it doesn’t generally work at the tail end of a tripling of prices that has left the market historically expensive on pretty much every measure available.

That said, Shiller’s trend-based argument is flawed.  The historical trend growth rate of real S&P 500 EPS that he uses–1.5%–does not apply to the current market.  The reason is straightforward, and centers on the impact that secular changes in the dividend payout ratio have had on growth.  EPS in the current market grows at a much faster pace than it did in the past, because a much larger share of current profit is devoted to growth-generating reinvestment, with a much smaller share devoted to the payment of dividends.

Now, here’s the problem.  We don’t have a clear, reliable, uncontroversial way to account for the impact that changes in the dividend payout ratio have had on earnings growth over time.  We are each then left to estimate the impact for ourselves.  Not surprisingly, those of us that want the market to go down estimate it to be a small impact, and therefore ignore the risk that it might undermine trend-based arguments.  Those of us that want the market to go up estimate it to be a large impact, and dismiss trend-based arguments altogether. Both approaches are wrong.

In this piece, I’m going to attempt to solve the problem.  I’m going to introduce a new type of EPS index, called the “Total Return EPS” index.  The Total Return EPS index represents what EPS would have been if corporations had never paid any dividends at all, but had instead used all of their dividend cashflows to buy back their own shares (or acquire or merge with other existing companies–there’s no difference in this context).  When regular EPS is converted to Total Return EPS, growth trends can be analyzed without concern for the distortive impact that changes in the dividend payout ratio have had over time, because the dividend payout ratio for all eras gets reduced to the same common denominator, 0%.

In previous pieces (on foreign share, financial share, technology share, sectoral balance of payments, and wealth redistribution), I’ve argued that corporate profits, though overextended, are not as overextended as they might initially appear to be.  A study of Total Return EPS confirms this argument. Regular S&P 500 EPS–shown in the Blodget/Montier/Brightman charts above–is a whopping 75% above its historical trend. But Total Return EPS is only 28% above its historical trend.

After introducing the Total Return EPS index, I’m going to teach the reader how to quickly and easily build it using Shiller’s publically-downloadable spreadsheet.  I’m then going to use it to redo his 2020 S&P 500 estimate, based on the data that was available in 2010.  In contrast to the bearish 1430 number that he arrived at, my estimate will come out to around 2150–still bearish, but not absurdly so, and significantly more likely to hit the mark, at least in hindsight.

It turns out that we can use Total Return EPS to do all sorts of interesting things: accurately predict future growth based on position relative to trend, decompose and visualize historical returns in terms of their contributing components, estimate profit margins during periods in the late 19th and early 20th century when the data necessary to calculate them was not available, construct new-and-improved Shiller CAPEs that allow for valid comparisons across history and across countries, and many more.  But those would be too much to discuss in one piece, so I’m going to save them for later.

Here, I’m simply going to introduce the concept, and use it to generate a new 2020 S&P 500 estimate.  Before I do that, I’m going to briefly outline the factors that have driven the secular decline in the dividend payout ratio.

Secular Decline in The Dividend Payout Ratio

An early dividend announcement from the Bank of North America.  Chartered in 1781 at the urging of Alexander Hamilton, the Bank of North America was the first Central Bank of the United States. It merged to form Wachovia Bank, and now exists as a part of the present day Wells Fargo ($WFC).

The market’s dividend payout ratio, which is the percentage of earnings that goes to dividends rather than to growth-generating reinvestment, has declined significantly over the last century.  In the chart below, I use the “Shillerized” version of the dividend payout ratio to illustrate the point.

shillerdpr

Why has the dividend payout ratio fallen so much?  We can point to at least three interconnected reasons:

First, the investment world is not as endeared to the idea of receiving dividends as it used to be, and therefore corporations are not under as much pressure to pay them.

“Do you know the only thing that gives me pleasure?  It’s to see my dividends coming in!” — John D. Rockefeller, quoted by John Lewis in Cosmopolitan, 1908.

Shareholders are perfectly comfortable, and often prefer, to see their idle excess funds deployed into other accretive activities–business expansion, share buybacks, acquisitions, mergers, and so on.  The increased liquidity and efficiency that the modern market affords ensures that any wealth created in these activities will immediately accrue to the benefit of shareholders in the form of higher stock prices.  In earlier periods, that was not necessarily the case.

Second, SEC rule 10b-18, originally issued in 1982, gave corporations that engage in share repurchases “safe harbor” from claims of market manipulation, removing a key legal risk that had otherwise discouraged the practice.  With that risk removed, share buybacks became a virtual no-brainer, an efficient and perfectly legal way of distributing capital to shareholders without causing them to incur unwanted tax liabilities.  When share buybacks displace dividend income, the money that continuing shareholders would otherwise have had to pay in dividend taxes stays inside the asset, where it is able to compound over time.

But what about situations where the market is expensive? Shouldn’t they encourage a shift away from buybacks, and towards dividends?  No.  Given the way that most people invest–with the “dividend reinvestment” option checked off–any dividends that are paid in lieu of buybacks will lead to the same outcome.  They will be used to purchase shares at the same expensive market prices–it’s just that in the dividend case, the purchases will take place outside the company, rather than inside.  The only relevant difference will be in the tax, which will be paid now rather than later.

Third, increased reliance on stock options as a form of performance-based employee compensation has created obvious reasons for corporate managers to prefer internal reinvestment to dividends.  When corporate managers pay dividends, the wealth contained in those dividends leaves the company, and no associated EPS growth is generated from it. Their claim on the wealth, which they hold through their stock options, is therefore lost, and any compensation they would have received for the EPS growth that the wealth would have produced, they do not receive.  But when they reinvest the wealth internally, it stays inside the firm, preserving their stock-option-based claims on it, and creating a direct boost to EPS that brings them that much closer to their performance targets.

It should come as no surprise then, that buybacks, acquisitions, and mergers now dwarf dividends as a destination for excess funds.  The following chart, borrowed from Michael Mauboussin’s fantastic piece on capital allocation trends in the U.S. corporate sector, illustrates the phenomenon.  The amount of capital recycled into mergers, acquisitions, and share repurchases is now roughly six times as large as the amount of capital returned to shareholders in the form of dividends.  100 years ago, that amount would have been an imperceptible fraction of the dividend amount.

maubassin

Constructing the Total Return EPS Index

According to S&P corporation, trailing twelve month (ttm) reported EPS for the S&P 500 is $102.77.  What would it be today, if, starting in 1871, all of the dividends that were paid out to shareholders had instead been used to repurchase shares (or acquire or merge with other companies)? That’s the question that the Total Return EPS index is trying to answer–not only for today’s date, but for all dates in market history.

To get an answer, we start with Shiller’s familiar spreadsheet, which contains average monthly prices and reported earnings for the S&P 500 and its pre-1957 ancestry, with data obtained from S&P corporation and the Cowles Commission.  We can use this data to create a log chart of EPS over time.  The EPS is shown below in red, with the exponential trendline shown in black and recessionary periods shaded in gray.  Recall that exponentials look linear on a log scale.

spxreg

This chart is essentially the Blodget/Montier/Brightman chart shown earlier.  As we see in the chart, current EPS is way above trend–by almost 75%.  If, right now, it were to fall back to trend, it would have to fall all the way down to $60.  At current prices, the index would then be valued at roughly 35 times earnings.

Now, let me make two brief points about the chart:

First, the chart shows inflation-adjusted EPS.  Every earnings data point charted in this piece is inflation-adjusted to today’s dollars.  When studying equity markets over the long-term, we absolutely have to inflation-adjust.  To not inflation-adjust would be to retain a source of noise and volatility in the data that we cannot accurately model or predict. Any patterns that we subsequently identify would automatically be called into question as potential cases of coincidence-exploitation and data-mining.

Second, the massive drops seen in 2003 and 2009 were the result of the application of accounting standards that were not applied to prior eras and that do not reflect true earnings performance.  In a perfect world, we would correct the chart accordingly, replacing the drops with more accurate estimates of earnings for the affected periods.  In some contexts, we will make such a correction, substituting operating EPS for reported EPS.  But in other contexts, we won’t, because it’s not necessarily needed, and will create false grounds for bears to dismiss what we’re trying to show.

In terms of analyzing the current equity market, it’s not crucial that we correct for the drops, because we’re not looking at Shiller-type averages of past EPS.  We’re only looking at trailing twelve month (ttm) EPS.  From the perspective of the ttm period ending in February 2015, the accounting-driven implosions of 2003 and 2009 have long since dropped out of the metric.  That said, when you look at the chart and attempt to glean patterns and trends from it, remember that the large drops in those periods are visual distortions, and are not supposed to be there.

The following chart shows regular EPS with the 2003 and 2009 periods corrected.  A mix of Bloomberg’s and S&P’s operating EPS series is inserted in the place of reported EPS starting in January 1998.

regulareps

If you look closely, you will notice that the EPS values in this chart and the uncorrected chart fail to adhere to a consistent long-term growth trend.  From the 1870s all the way through to the 1930s, there was essentially zero EPS growth.  From the 1930s through to the 1990s, there was more EPS growth, but still not all that much. Then, starting in the mid 1990s, EPS growth exploded, achieving in a 20 year period what had earlier taken 60 years to achieve.  How can we fit all of these periods onto the same trendline?  We can’t, which is why the chart is ugly and uninspiring–a straight line forced onto a highly malformed time series.

To turn the time series into something more visually compelling, we need to build an index that puts all periods of history on an equal footing with respect to the tradeoff between dividend income and earnings growth.  If we do that, everything will fall into place.

Enter the Total Return EPS Index.  The Total Return EPS Index models what EPS would have been if the dividend payout ratio had been 0% at all times in history, with all dividend cash flows instead used to repurchase shares (or acquire or merge with existing companies).  The Total Return EPS index reduces all periods to that common denominator, so that different periods can be accurately compared with each other, and a true underlying growth trend legitimately extracted.

Now, there’s one crucial, counter-intuitive caveat that we need to insert.  The index should not assume that shares are bought back at market prices.  For if it makes that assumption, then periods of history where stocks were cheap will show abnormally high growth, driven by the high rate of return that the share buybacks, conducted at cheap prices, will have produced.  Conversely, periods of history in which stocks were expensive will show abnormally low growth, driven by the low rates of return that the share buybacks, conducted at expensive prices, will have produced. We’re trying to determine the historical trend in EPS growth.  Valuation–the cheapness and expensiveness of stock prices in any given period–has nothing to do with that trend.  We therefore need to remove its effects from the index.

Instead of assuming, as most “total return” indices do, that shares are bought back at market prices, our new index will therefore assume that they are bought back at prices that reflect the same underlying value across history.  Such an assumption is needed to ensure that all periods of history are placed on an equal footing, to allow for accurate comparisons between them.

But how shall we measure value across history?  The best available tool–far from perfect, but adequate to the task–is the Shiller CAPE.  We’re going to assume, then, that shares are bought back at prices that correspond to a constant Shiller CAPE across history–in this case, a Shiller CAPE equal to the historical (geometric) average of 16.59, which we will arbitrarily refer to as “fair value.”

Unfortunately, for the 1871-1881 period, there aren’t 10 years worth of trailing data on which to calculate a Shiller CAPE.  So, for that period, we’re going to have to use the simple ttm P/E ratio.  We will assume that shares are bought back at a ttm P/E ratio equal to the historical (geometric) average of 15.54.

Now, to the spreadsheet itself.  Here’s the concept.  We start by creating a column that contains an arbitrary starting share count. We make the initial share count, in the first month of the series, equal to 1.0 (or whatever number you want).  We then shrink the share count each month by the amount of shares that the monthly dividend would have been able to purchase at fair value prices (again, “fair value” as measured by the Shiller CAPE for periods after January 1881, and the ttm P/E ratio for periods before January 1881.)

The images below show the technique in practice from 1871 to 1881.  In each month, we shrink the share count by the amount of shares that the dividend for that month would have been able to buy, assuming a fair value ttm P/E multiple of 15.54.

a3kla

In January of 1881, the first month where a Shiller CAPE reading is available, we shift to that measure of fair value.  In each month, we shrink the share count by the amount of shares that the dividend for the month would have been able to buy, assuming a Shiller CAPE of 16.59.

sadker

What we end up with is a gradually contracting share count that tracks what would have happened to the actual share count if all dividend payouts had instead been used to repurchase shares.  Dividing the actual reported EPS by that share count, we get the Total Return EPS, which is what the EPS would have been under a 0% dividend payout ratio.

dafaaea

But if all dividends had instead been used to repurchase shares, then the market’s price on the subsequently higher earnings would have been higher. We therefore need to calculate a new price index, a price index that reflects what the price would have been, on the assumption that the market had valued the now-higher earnings in the same way, with the same P/E multiple.  To do this, we divide the actual price by the contracting share count, just as we divided the actual EPS by that share count.  We end up with the Total Return Price Index, which goes together with the Total Return EPS index (we will use this price index in a future piece to calculate a new-and-improved Shiller CAPE).

aefvv

The following chart plots the uncorrected Total Return EPS on a log scale.  At present, it’s roughly 28% above its historical trend, versus 75% for the uncorrected regular EPS.

spxtrepske

The following chart plots the corrected Total Return EPS on a log scale.  At present, it’s roughly 21% above its historical trend, versus 58% for the corrected regular EPS.

adjopeps

Notice, in particular, the dramatic improvement in the consistency of the trend.  The growth from the 1870s to the 1930s, for example, is on par with the growth from the 1930s to the 1990s, which is on par with the growth from the 1990s to now.  In the charts of regular EPS shown earlier, that was not the case.  The first period, when the dividend payout ratio was high, had virtually no growth; the second period, when the dividend payout ratio was lower, had moderate growth; the third period, when the dividend payout ratio was lowest of all, had substantial growth.

Annualized, the historical trend in Total Return EPS growth comes out to around 5.5% per year for the uncorrected case, and 5.7% per year for the corrected case–both close to the generic 6% number that represents the U.S. corporate sector’s historical return on equity.  This is not a coincidence.

A New 2020 Estimate

I’m now going to redo Shiller’s 2020 price estimate using the corrected version of Total Return EPS.  As of June 2010, nominal EPS was $69.43.  Total Return EPS was 7.5% below its historical trend.  Boosting nominal EPS by an amount sufficient to bring Total Return EPS back to trend, we get $75.  So, if EPS in 2010 had been $75, Total Return EPS would have been perfectly on its trend.

The Total Return EPS, as constructed, grows at an average rate of 5.7% per year. Inflation adds roughly 2%, and the dividend, using the yield at the time, subtracts roughly 2%.  So we get 5.7% + 2% – 2% = 5.7% as the expected nominal EPS growth rate. Applying that expected growth rate out over a 10 year period to the starting $75 number, we end up with $131.  That’s the EPS estimate that this trend-based method produces for the year 2020. Multiply by 15 and we get 1965, versus Shiller’s overly bearish 1430.

Now, the 15 number that Shiller used as the “average” ttm P/E ratio represents the average ttm P/E ratio from 1890 to 1990.  An average from a later, more relevant slice of history would probably be more appropriate.  The (geometric) average ttm P/E ratio from the S&P 500’s inaugural year of 1957 to present is 16.46.  Using that number, we arrive at a year 2020 S&P 500 price estimate of 2156 (which, interestingly, is roughly the same as the year 2020 estimate–2154–that I arrived at using a more direct method in a piece from 2013.)

These estimates assume a reversion of both earnings and P/E back to their properly-measured historical trends.  Such a reversion may not occur.  But if it were to occur, it would not be anything historically extreme–just a drag on EPS growth that brings final EPS to a value ~15% to ~20% below where it would be if it grew at the trend rate from today forward, coupled to a downshift in the P/E ratio from the current 18+ back to the average of around 16.5.

It may be hard to believe that such a reversion could ever take place in this teflon, never-lose, never-give-a-bear-a-damn-thing U.S. equity market, but go ahead and #timestamp that estimate, we’ll come back to it.  A number of impeding forces are likely to push back on U.S. equity performance between now and 2020, that have only recently started to rear their heads.

Posted in Uncategorized | Comments Off on Introducing the Total Return EPS Index: A New Tool for Analyzing Fundamental Equity Market Trends

Janitor to Multimillionaire? Not In This Market

Last week, CNBC profiled the inspirational story of Ronald Read, a gas station attendant and janitor from Vermont who amassed an $8MM fortune simply by investing portions of his small salary into high-quality, dividend-paying U.S. equities.

Judging from the familiar names that Read is reported to have owned–AT&T, Bank of America, CVS, Deere, GM, GE, and so on–alpha generation doesn’t appear to have been the main driver of his success (though it was probably a contributor).  Over the last 40 years, those names have roughly tracked the large cap index, as you can see in the chart below.

read

The main driver of Read’s success appears to have been discipline: saving diligently and investing efficiently, and sticking with the process over a lifetime.  That’s good news, because it means that anyone can potentially do what Read did.  No special stockpicking talent is required.

Would it be possible for a present-day investor in Read’s employment circumstances to amass the fortune that Read amassed, if that investor were to apply Read’s discipline to a U.S. equity strategy?  The actual numbers were crunched, and the surprising conclusion was yes.

cnbcread

Per the analysis, if an investor wanted to amass an $8MM fortune, she would need to put around $300 per month into the market.  Compounded at an 8% rate over 65 years, the ensuing investment pile would be worth around $8MM today.  Now, $300 per month is a lot of money for someone working a minimum-wage job.  But it’s not a prohibitive amount. $10 per hour times 40 hours per week times 4 weeks per month is $1600.  On a pre-tax basis, $300 is roughly 20% of that amount.

But there’s a mistake in this analysis.  It makes no adjustments for inflation. Sixty-five years ago, $300 per month was a very large amount of money–150% of the monthly wage of the average non-supervisory worker, which was itself almost double the federally legislated minimum wage.  No janitor on earth would have been able to afford it.

Now, to be fair, the 8% return assumption used in the analysis might refer to a real, inflation-adjusted return, rather than a nominal return.  If so, then the analysis is based on a return assumption that will almost certainly prove to be wrong.   There have only been a few periods in U.S. history where a long-term buy-and-hold investor could have earned 8% after inflation.  Needless to say, those periods–marked by widespread skepticism and obliviousness towards equities, and therefore extremely cheap equity valuations–had very little in common with the current market environment.

The following chart shows the annualized real total return of the S&P 500 for 65 year periods beginning in 1871 (endpoint 1936) and culminating in 1950 (endpoint today). All dividends are reinvested at market prices.  The 1932 bear market low is circled in red for reference.

spx33

As you can see, the real total return only breached 8% on two occasions: first, during the troughs of the Great Depression (specifically, from April of 1932 into 1933, and then after the 20% correction that ended in September 1934, which took the index back to its January 1932 level), and second, during the selloff that occurred in the months after the Pearl Harbor attack (with a final bottom in April of 1942, which again took the index back to its January 1932 level).  The market’s valuations at the lows of those periods, as measured by the Shiller P/E ratio, were 5.5 and 8.5 respectively, versus 27.2 today. Notably, in the case of the Great Depression, the return over the subsequent 65 year period was artificially boosted by the valuation extreme seen at the endpoint–1934 plus 65 years conveniently equals 1999 (party like it’s).

To get a better answer, let’s redo the analysis.  How much would an investor have to put aside each month to have $8MM at the end of a long investment lifetime, one that begins today, say, at age 25, and that ends at Ronald Read’s final age of 92?  Rather than apply an arbitrarily chosen annualized real return assumption (and then argue over whether the assumption will prove to be accurate going forward), we can just run the simulation directly, using the numbers that the market actually produced during Ronald Read’s investment lifetime.

What we’ll need to do, then, is calculate the amount, in 2015 dollars, that an investor would have had to have put into the S&P 500, each month starting in 1948, to end up with $8MM today.  Comparing that amount to the current salaries of low-wage workers, we will be able to see whether or not the strategy is realistically affordable.

To have produced an investment pile worth $8MM today, on the assumption that all investments occur in a tax-sheltered account, with no transaction fees and no management fees, an investor would have had to have put aside $524 each month in 2015 dollars.  $524 is a little over 30% of the gross monthly minimum wage.  The investor would therefore have had to have found a way to live on a bit less than $1100 per month, before taxes.  Such a frugal feat would have been extremely difficult to accomplish; in the typical real-life situation, where there are actual mouths to feed, it would have been impossible.

And note that taxes are not negligible here.  Even if we assume a 0% income tax (netting out the income transfers that the worker would be eligible to receive), the combined 7.65% that the worker would have to pay in social security and medicare taxes (on the gross wage) would reduce the after-tax, after-investment monthly income to a number below $1000.

Of course, as we move up the income chain, the prospects get brighter.  The average hourly wage for non-supervisory workers in the U.S. is around $21, more than double the federally legislated minimum wage.  $21 per hour comes out to around $3400 per month. Taking out income and payroll taxes, and assuming no income transfers received, the after-tax income falls to around $2800.  It’s reasonable to think that a determined worker could manage to take $524 out of that amount each month.  And so even if the numbers don’t work out for janitors and gas attendants, they might still work out for average employees.  The redemptive, feel-good conclusion that motivated the profile is therefore preserved: that Wall Street can, in fact, make average people, who have no special stockpicking talents, into multimillionaires.

But there’s another problem.  In order for present investors to be able to earn the kinds of returns that would transform contributions of $524 a month into an $8MM sum over a 65 year period–the kinds of returns that the market actually saw from 1948 until today–the market will have to do that dirty thing that it did multiple times from 1948 until today, the thing that those of us who are invested in it pray for it not to do: CRASH.

It’s true that $524 per month invested in U.S. equities from 1948 to present would have produced an investment pile worth $8MM.  But those monthly contributions, and the quarterly dividends earned along the way, would have been invested and reinvested into the market at the ultra-cheap valuations that the many crashes, corrections and bear markets seen from 1948 onward produced.  Make no mistake: those valuation depressions were absolutely critical to the attractive returns that investors employing the strategy would have earned–and that Ronald Read did earn. Without them, the returns to the strategy would have been far less impressive.

Right now, we are not anywhere near the kinds of depressed valuations that helped to power the outsized returns of the post-war era, not even on measures that the most extreme of market optimists might otherwise embrace.  And so, unless the market sees meaningful downside (or long, frustrating water-treading) in the years ahead, downside (or water-treading) that makes stocks genuinely cheap again, investors employing Read’s discipline should not expect to achieve his lucrative outcome.

What if stocks from 1948 to 2015 had never crashed, never corrected, never entered ugly bear markets that depressed their valuations?  What if their valuations had instead always equaled today’s valuation, 27.2 on the Shiller P/E?  How much would an investor’s monthly contributions have had to have increased, to have produced the same $8MM in the final tally?  It turns out that we can give an exact answer to this question.

We know what the S&P 500’s earnings were at each point in time from 1948 to 2015.  We can therefore calculate what its prices would have been at all times during the period if its valuation had stayed constant at today’s level, a Shiller P/E equal to 27.2.  We can then calculate what size of contribution would have been required to produce the same $8MM investment pile, assuming that the contributions had been invested, and the dividends reinvested, at those hypothetical prices.

The following chart shows the actual S&P 500 price index alongside a hypothetical price index which postulates a Shiller P/E ratio that always equals the current value–27.2. Note that the y-axis is log scale, and all prices are nominal.

actualhypothetical

As stated earlier, to end up with $8MM by making regular investment contributions (and reinvesting dividends) into the blue line (the actual S&P 500 over the period), an investor would need to commit $524 (in 2015 dollars) per month.  To end up with the same fortune by making regular investment contributions (and reinvesting dividends) into the red line (the hypothetical S&P 500 that never crashes, whose valuation is always equal to today’s valuation, a Shiller P/E of 27.2) an investor would need to commit $1831 per month–almost four times the earlier amount!

The average non-supervisory worker, who makes $2800 per month after tax, cannot afford to plow $1831–65% of her net income–into the stock market.  And so unless stock valuations are going to retreat markedly from current levels (or spend very long periods of time going nowhere), investors shouldn’t expect to be hearing about Ronald Read stories in the next go-round.

Now, the Shiller P/E might not be the best way to measure valuations in the current market.  But even on valuation measures that are more generous, the simulation still points to a substantial increase in the required contribution.  Per S&P Capital IQ, the current trailing twelve month P/E ratio on operating earnings is around 18.5. To have generated $8MM from a market that always held at that valuation, an investor that started in 1948 would have had to have contributed $1005 per month–roughly double the earlier amount.  And note that in using a one year earnings measure to gauge valuation, we are implicitly assuming that profit margins over the next 65 years will average out to their current record-high levels–possible, but a very risky assumption to make.

In all likelihood, present and future valuations will prove to be more important to returns over the next 65 years than they were to returns over the last 65 years.  The reason why is that the growth of the population is set to slow, with the average age set to increase significantly.  Slowing population growth and a significant upward shift in the average age, towards the elderly, implies reduced aggregate demand growth, and therefore a reduced need for expansive corporate investment.  Corporations, if they want to do well for their shareholders, will therefore have to shift their capital allocation strategies away from traditional capital expenditures towards what might be perjoratively described as “capital recycling”–the payment of dividends (which get reinvested into the market) and the conduct of share buybacks (which are essentially identical to reinvested dividends in terms of their effect on total return).  But the rate of return that dividends (reinvested at market prices) and share buybacks (repurchased at market prices) produce is strongly dependent on the valuations at which the reinvestments and repurchases occur.  It follows that as the corporate sector shifts towards capital recycling as an allocation strategy (a shift that is already well underway), valuations–not only at the moment of purchase, but also during the entirety of the holding period–will become increasingly important to the market’s return prospects.

To illustrate the point, imagine a demographically and technologically stagnant future world where dividends are punitively taxed, and where capital expenditures, in excess of depreciation, are neither needed nor profitable–a world where such expenditures do nothing but fuel competition, deflation, and profit margin shrinkage–put simply, a world that is Japan.  If corporate managers in such a world are good stewards of capital, they will deploy 100% of their earnings into share repurchases. EPS growth will then be entirely determined by how much the repurchases contract the S, the share count.  But the amount by which a given repurchase event contracts the S, the share count, is determined by the repurchase price–and therefore, the valuation.  And so, in such a world, expensive valuations will depress EPS growth, and by extension, investor returns.  The simulation above bears this out, with reinvested dividends as the proxy for share buybacks.

Crucially, a large portion of the EPS growth that was realized over the last 65 years was the result of expansive capital formation, the net building of new things that produced new profits, and an EPS that grew by the E.  In a futuristic Japanese world where that driver of EPS growth is removed, and where all EPS growth results from repurchases, a shrinking S, the returns to an always-expensive scenario will be that much weaker.

If you’re an investor with a short time horizon, you should want valuations to stay high, or even better, go higher, into a bubble, so that you can get the most out of your holdings when you cash them out.  But if you’re a disciplined investor that is in this for the long term, particularly a 20-something, 30-something, or even early 40-something, with a lot of income yet to be earned, you should not want valuations to stay where they are.  You definitely should not want them to go higher, into a bubble.  Instead, you should want the opposite of a bubble, a period of depressed valuations–the lower the better.

Granted, a rapid downward move in the markets, towards valuations that are genuinely cheap, would entail the pain and regret of mark-to-market losses on present holdings.  But that pain and regret will only be short-term.  In 20 or 30 or 50 or 65 years, the paper losses, by then evaporated, will have been long since forgotten, having proven themselves to have been nothing more than opportunities to compound wealth–monthly contributions, reinvested dividends, and share buybacks–at high rates of return.  As Ronald Read’s example shows, the compounding adds up over time, allowing disciplined investors–even those of modest means–to build surprising fortunes.

Posted in Uncategorized | Comments Off on Janitor to Multimillionaire? Not In This Market

Technology and Finance: Drivers of a Profit Margin Explosion

In this piece, I’m going to show that the profit margin expansion seen in the U.S. corporate sector over the last two decades has been driven largely by gains in the financial and technology sectors.  I’m then going to examine arguments for and against the sustainability of this shift.

Profit Margin Contributions By Sector

The following chart shows the aggregate net profit margin of publically-traded U.S. equities with market capitalizations greater than $200MM from January 1964 to October 2014.  Recessionary periods are shaded in gray.

netprofm

The next chart separates out the contribution to the aggregate profit margin by sector. For each sector, the colored area represents the individual earnings of the sector divided by the total revenues (sales) of all sectors.  Note that the sum, the black line, is just the aggregate profit margin shown in the previous chart.

EPSshare

Notice the rising contribution from the financial sector (light purple) and the technology sector (olive green), and the falling contribution from the other sectors in aggregate:

dklskls

spexspe

In January of 1964, financial and technology sector earnings contributed 0.49% to the aggregate profit margin, which was 6.60% at the time.  Today, they contribute almost seven times that amount, 3.42%, to an aggregate profit margin of 8.09%.

Changes in  Sectoral Revenue Contributions

A better way to think about what has happened here is to think in terms of sectoral revenue contributions: revenues of individual sector as a percentage of the total revenue of all sectors.  The following chart shows the evolving revenue contributions of each sector, from January of 1964 to October of 2014.

salesshare

As the chart illustrates, the revenue contribution of the combined financial and technology sectors–the amount of total revenues that are revenues from those sectors–has increased substantially over time.  In January of 1964, the revenue contribution was 5.41%.  Today, it is 23.83%–almost a quarter of the total.

E4A3A

DASDA

This change in revenue contribution matters because technology and financial sector revenues tend to be earned at higher profit margins than the revenues of other sectors: historically, 7.74% for technology and finance, versus 5.19% for the rest.  And so if the revenue contribution from the financial and technology sectors has increased, then we should expect the “normal” profit margin of the aggregate corporate sector–if there is such a thing–to have increased as well.  The targeted “mean” in a “mean reversion” will have shifted upward, rendering the ensuing picture less bearish.

(Note: the revenue contribution of healthcare, shown in brown, has also increased substantially over time.  But, to the surprise of many, current healthcare profit margins, at 6.2%, are below their own historical average, and only slightly above the historical average of the aggregate corporate sector.  They are not appreciable contributors to current profit margin elevation).

It turns out that we can correct for this shift, creating an “adjusted” profit margin that accounts for the effects of changing sectoral revenue contributions.  What we need to do is take the average historical revenue contribution of each sector, and compute what the aggregate profit margin would have been, at each point in time, if each contribution had been equal to its individual historical average.

The following table shows the historical average revenue contributions of each of the 10 GICS sectors:

fa3a

So, historically, energy revenues have represented 12.78% of total revenues, materials revenues have averaged 7.80%, health care revenues have averaged 5.46%, and so on. What we want to know is, what would the profit margin of the aggregate corporate sector have been at each point in history if the revenue contribution of each individual sector had been equal to its average?  This “adjusted” profit margin will filter out changes that have been driven solely by shifts in sector size and contribution, and will thus provide a more accurate picture of the aggregate profit margin to use when making historical comparisons.

At this point, Bulls are probably hoping that I pull out a chart showing that when profit margins are properly adjusted in this way, that they end up not being historically elevated. Sorry, not quite.  As the chart below shows, the adjustment doesn’t make much of a difference.

AA3A

The unadjusted profit margin (blue) is 45% above its historical average, versus 35% for the adjusted profit margin (green).  Relative to the respective averages, the adjusted profit margin is only about 10% “less elevated” than the unadjusted profit margin.  This difference is worth something, no question–but it’s not enough to eliminate profit margin concerns outright.

The reason that the adjustment doesn’t make the kind of difference that we might otherwise expect is that the profit margins of the financial and technology sectors have themselves expanded dramatically in recent years.  The truth is that profit margin increases within the financial and technology sectors, rather than increases in their contribution to total revenue, have been the primary drivers of the aggregate profit margin increase.

The following chart shows the profit margins of the combined finance and technology sectors (red) alongside the profit margins of the combined other sectors (blue):

A3SES3A

As you can see, profit margins in finance and technology have exploded.  Combined, they are running at almost twice their historical averages–86% above, to be precise.  The profit margins of the combined other sectors are hardly elevated at all–only around 18% above their historical averages.

The following charts show the profit margins of the finance and technology sectors individually:

A33SE3A3

DFA4FRG

Notice that a large chunk of the move is recent–a phenomenon unique to this specific cycle–especially in the technology sector.

Explaining the Rise

It goes without saying that finance and technology, which together represent over 42% of current U.S. corporate earnings, are two sectors that we should keep a close eye on going forward.  Changes within them have driven the profit margin expansion of the last several years, which itself has driven the bull market, having made possible a “goldilocks” scenario in which earnings have been able to grow robustly despite slow top-line growth and almost non-existent inflation.  The slow top-line growth and almost non-existent inflation has pushed the Fed into an aggressively easy monetary stance that has served as fuel for persistent P/E multiple expansion, with more and more investors ditching the misery of zero-yield cash and bonds to join the market advance.

The finding that the profit margin expansion has been driven largely by changes inside the finance and technology sectors sheds doubt on other stories that have been offered as explanations.  Weaker labor unions, increased access to cheap foreign workers, a rise in earnings taken in from abroad, lower corporate taxes, more effective corporate tax avoidance schemes, and so on–these explanations fail to make sense of the fact that profit margins haven’t increased nearly as much in sectors outside of finance and technology. Whatever the correct explanation for the current state of profit margins ends up being, it needs to be one that applies with some preference to finance and technology, which is where the most dramatic shift has taken place.

What, then, is the explanation for the rise?  Why have profit margins in finance and technology increased so dramatically over the last several years?  Will the increase hold up?

A Bullish Angle

On the finance front, bulls can make a compelling argument that the financial sector’s contribution to the profit margin increase is likely to be sustained.  The increase in the financial sector share of total revenues has been driven by higher debt levels across the economy–that change will almost certainly prove to be secular.  At the same time, the increase in profit margins within the financial sector has arguably been driven by the drop in short-term interest rates (funding costs for financial institutions), which is a change that is also likely to be secular.  Note that the last time that financial profit margins were at their current levels was in the early 1960s, when short-term interest rates were low.  The Fed tightening cycle that lasted from the late 1960s through the 1980s seems to have been what pulled them down, as they fell much more precipitously during that period than the profit margins of any other sector.  They only began to regain their prior levels in the mid-to-late 1990s, as the Fed shifted to an easier monetary stance.

On the technology front, bulls can make a similarly compelling argument that the revolutionary technology of the information economy, which has only been fully fleshed out in the last decade, has been the game-changer, having created an increasingly “winner takes all” environment in which it has become more and more difficult for potential new entrants to credibly compete with the first-mover.  If they are right, then profit margin mean reversion–in the technology sector and in any other sectors that manage to piggy back on the dynamic–would seem to be less likely.

A Bearish Angle

Of course, bears can offer compelling counter-arguments to this optimism.  On the finance front, they can point to the fate of the yield curve–which, assuming the Fed follows through on its normalization plans, does not look good.

When the Fed cuts rates for the first time, the long end of the curve usually stays put. People continue to expect an eventual return to normalcy, and price the long end accordingly.  The result ends up being a steep curve that boosts financial sector profit margins.  But when the Fed cuts rates and keeps them cut, for a period that seems to drag on forever, because the economy never seems to get hooked into the kind of genuine inflationary expansion that would justify a tightening cycle, the market eventually figures things out. Investors realize that long-term rates need to be lower, and pulls the long-end down accordingly, at the expense of financial sector profitability.

Eventually, the Fed will raise the short-end–if not simply out of a desire to restore some normalcy to monetary policy.  When that time comes, the long end will again be slow to respond–this time slow in the opposite direction, slow to rise, given the anchoring and inertia of market participants who, by then, will have grown accustomed to the idea of secularly low interest rates.  The result will be a yield curve that gets flatter and flatter with each hike, and a financial sector whose profit margins get squeezed.  That seems to be exactly where we are currently headed, and it is not bullish.

In the most recent earnings data from banks, we’ve seen a consistently weak trend: flat YOY revenue growth and negative YOY EPS growth, brought on by increased competition, particularly among smaller banks, increased regulatory expenses, and reduced profitability due to a flattening yield curve.  Loan growth, which would otherwise represent the bright spot, is not making up for the reduced profitability.

On the technology front, bears can make a similarly compelling argument.  “Tech” is the most cutthroat and competitive of all sectors.  Historically, it has produced subpar returns for investors (ranked number 6 out of 10 sectors), likely due to the way in which disruption and competition have worked to break down dominant positions within it. When we look at the seemingly impenetrable empires of the $AAPLs and $GOOGs and $MSFTs and $FBs of the world, it can be tempting to think that the truly massive levels of profitability they currently enjoy will be forever secure–but this kind of thinking is not supported by history.

There’s a particularly interesting and relevant analogy that bears can raise in this context, one that involves a different sector: commodities.  The historical evidence on the real return potential of spot commodities is overwhelming: there is no real return potential, spot commodities do not offer real returns.  For proof, consider the following 130 year chart from Dylan Grice.  Notice the black line languishing stupidly at the bottom:

realcommods

But then again, over the last decade, we saw a massive boom in commodity prices around the world.  As always happens, compelling stories emerged to explain why the boom had occurred and why it would almost certainly hold up–insatiable demand growth from China, India, and other emerging markets, an increasingly constrained supply that fails to grow, even in response to large price increases, and so on.  If you had told people in 2007 or 2008, or in 2010 or 2011, that these were just stories, and that there would eventually be a painful reversion to the mean just a few years later, very few people would have taken you seriously.  Nobody in the commodity complex at the time was seriously entertaining the possibility.

But now here we sit, in a healthy economic expansion, with oil trading below $45 (!!), the same inflation-adjusted price that it traded at 30 years ago, near the lows of the last oil downturn.  A 30 year period of zero real returns for this and other spot commodities has once again vindicated the apparent lesson of history: that spot commodities do not produce real returns.  Now, to be clear, I don’t expect profit margin bearishness to receive the same degree of vindication–but some caution and humility are certainly in order, given the possibility.

Posted in Uncategorized | Comments Off on Technology and Finance: Drivers of a Profit Margin Explosion

Intrinsic Value: Interest Rates, Inflation, and the Forgotten Concept of the Time Value of Money

In the previous piece, I offered a definition of the investment concept of “intrinsic value.” Intrinsic value is the value that the owner of a security realizes from holding the security, rather than selling it.

To determine the intrinsic value of a given security, we can apply a simple test.  We posit that the security that cannot ever be sold, but must be held until maturity.  We then ask ourselves: what is the maximum price that we would be willing to pay, or alternatively, the maximum amount of cash that we would be willing to exchange, to own the security?  That amount of cash must equal the intrinsic value of the security, the value that accrues to us simply from owning it, otherwise the exchange would not make rational sense.

In this piece, I’m going to explore the set of fundamental considerations that would impact a rational agent’s assessment of the intrinsic value of different types of securities.  The analysis will seek to clarify “the way things ought to be” in financial markets–the way they would be if everyone invested rationally, based solely on the intrinsic value contained in the investment opportunities presented.

To be clear, “the way things ought to be” in financial markets is not the way things actually are, particularly with respect to long-dated assets, assets whose maturities are too far out in the future to “wait for.”  Market participants that trade and invest in long-dated assets do so based not on estimations of “intrinsic value”, but rather on estimations of how the prices of those assets will evolve over the short-term to medium-term, a few months to several years, which is the limit of human look-forward capacity, and the time horizon on which investor performance is measured.  Investors are not able, personally or professionally, to seriously consider longer time horizons, on the order of decades or even centuries, even though that is often how long it takes for the “intrinsic value” of long-dated assets to play out.

Investors worry about the “fundamentals” of long-dated assets not for their own sake, but because the fundamentals influence the prices, through non-fundamental perceptual and behavioral channels.  The fundamentals serve as subjective inputs into the minds of investors, factoring into the rule-based calculations that drive actions and outcomes in the market: “X is happeningit probably means Y.  From a portfolio standpoint, the right move is probably for us to do Z.”

Cash, Bonds, Stocks, Other

Investors are confronted with a range of different types of assets in which to hold their wealth.  We can simplify this range into four categories: Cash, Bonds, Stocks, and Other. To determine the “intrinsic value” of assets in each category, we need to express them in terms of cash, which is the basis for measurement.

Cash is just cash, money, whatever must be accepted by law to repay debts, public and private.  The prices of all assets are expressed in terms of it, therefore the intrinsic value of one unit of it is one.  Bonds are a finite collection of more-or-less guaranteed cash payments, usually consisting of small cash payments for a time (coupons), followed by a large cash payment at the end (return of principal).  Equities are an infinite collection of non-guaranteed cash payments (dividends, or rental payments on the use of capital, land, housing, and so on).  The “other” category consists of unproductive assets, assets that do not generate a cash flow–think, gold bars.  These assets have very little intrinsic value, and are almost always purchased with the intent of eventually offloading the investment onto someone else.

The dividends that shares of equity pay out to their owners tend to grow at a rate that exceeds the rate of inflation.  The reason is twofold:

  • First, the dividends are backed by corporate earnings, and are paid out as a percentage thereof.  On a unit basis, corporate earnings equal price minus cost.  Inflation–a change in the price index–acts to increase both of these entries equally, therefore it acts to increase their differenceearningsequally as well.
  • Second, not all of the earnings are paid out as dividends.  Some of the earnings are used up in the purchase of growth.  The amount of growth purchased adds additional earnings, and therefore additional dividends, to the numbers of the future.

Now, to be clear, funding costs–for example, dilution–can cause per-share earnings to not keep up with inflation, particularly when the corporate sector is inefficient in its use of the proceeds.  In the present context, we will assume that corporations fund their growth internally, without increasing share count (an assumption that has proven valid in recent experience), rendering the issue of dilution moot.

Leaving the “other” category aside, we are left with two types of assets whose intrinsic value we want to measure: bonds and shares of equity.  So we return to the critical question: how much of each type of asset would we be willing to exchange for cash, if we could not ever go back on the exchange?

A bond is just a future stream of small cash payments (coupons), followed by a final payment (repayment of principal).  How much cash, held in hand right now, would we be willing to trade for that future stream?  The answer, for us, is the intrinsic value of the bond.

A share of equity is a future stream of small, growing, inflation-linked cash payments, without a maturity date. How much cash, held in hand right now, would we be willing to trade for that future stream?  The answer, for us, is the intrinsic value of the share of equity.

The Time Value of Money

Money now obviously is not the same as money later.  Money now is better, for a myriad of reasons, not the least of which is that it can be used now, at the option of its owner. Money later, in contrast, cannot be used until it is received.

The difference between money now and money later is the “time value of money.”  Looking at the current state of bond prices in the developed world, we might think that money has no time value to current investors.  After all, long-term bonds in the developed world trade at prices with implied yields approaching zero.  An investor who lends his money to a government in the developed world for five, ten, twenty, even thirty years, gets essentially nothing in return–nothing except the original money that was lent out, for a net nothing.

But to conclude that money has no time value to current investors would be a huge mistake.  The reason that current investors are willing to lend their money to governments at zero rates of interest for prolonged periods of time is that they know they can easily get out of the loans by selling the associated securities back into the market.  For all intents and purposes, in a liquid market, where investors are confident that they will be able to sell their investments at or near cost, the “time value of money” loses relevance.  The assets become the functional equivalents of “money now”, given that they can be converted into “money now” at the push of a button.  In purchasing the asset, the investors don’t have to “part” with their money, therefore they don’t have to put a price on the cost, to them, of “parting” with it.  If they did, the price dynamics observed in developed world bond markets would be very different from what they are.

Take any institutional fund that is currently eager to lend its money to the Swiss or Japanese governments for decades at near-zero interest rates, and tell that fund that it will have to hold the associated debt security until maturity–that it will not, under any circumstances whatsoever, be allowed to get the money back by selling the security to other investors (or engaging in any “tricks” that might simulate a sale, such as posting the security as collateral for a loan, or selling short a similar security).  You would quickly see the time value of money come back into play, in a dramatic way.  To be sure, it is not zero.  Not even close.

When word gets out that a financial institution is in trouble and is facing a liquidity crisis, its customers rush to redeem their money.  The main fear that drives their behavior isn’t the prospect that the money will be lost forever–the investors know they will almost surely get it back, after everything has been cleaned up, years later.  Their most pressing worry is actually the prospect that the money will get stuck inside of a black hole in the interim–a confused, entangled “what belongs where?” scenario, a court battle involving complicated and drawn-out litigation–and that the customers therefore won’t be able to access the money for months, years, maybe decades.  Ouch.  Again, we see the importance of the “time of value of money”–when it is actually at stake.  In a stable, liquid market with a confident bid, it is not at stake, and so it doesn’t factor in.  But things can easily and quickly happen to put it at stake, which is why long-term assets–assets with maturities on the orders of many years, decades or centuries, that cannot realistically be “waited out”–are prone to violent bouts of volatility, when confidence in the presence of future bids near the current price is lost.

The Exercise

Cash held in the banking system carries essentially all of the benefits of cash held in hand, with a number of additional perks and conveniences.  For this reason, individuals usually choose to hold their cash in banks, in the form of deposits.  The banks normally pay interest on the deposits, which they fund through the income they generate on their loans. Without getting into the details, central banks in modern financial systems have the ability to adjust the rate of interest that banks, in the presence of market forces, have to pay on their deposits.  Expectations with respect to the future path of this rate of interest have a substantial effect on the “intrinsic value” of all other assets, because all other assets must compete with it.

To illustrate, let’s do the exercise.  You have $100,000 in wealth, and three modalities in which to store it:

(1) Cash: You can hold the wealth as a cash deposit in an insured bank, and earn interest on it.  While in the bank, you will be able to spend it on consumption.  Depending on the nature of your deposit, you may have to wait a few weeks to spend it, maybe a month or two, but you can afford to wait that long.  To keep things fair, let’s suppose that if you choose this option, you can only spend the money on consumption–you cannot go and invest it in the other options later.  You will have to make the “investment” decision now, and you will have to stick with it for good, at least as far as that money, the $100,000, is concerned.

The interest rate that you will be paid will be determined by the central bank, chosen so as to bring the rate of inflation–and any other macroeconomic target that the central bank might have–onto target.  In periods where there are strong inflationary pressures, the rate will be increased, so as to incentivize you and others to hold your cash rather than spend or invest, and to disincentivize others from borrowing it to spend or invest.  The same is true in reverse: in periods where inflationary pressures are weak or negative, the rate will be reduced, so as to disincentivize you and others from holding your cash, get you to spend or invest it instead, and to incentivize others to borrow it to spend and invest.

(2) Bond: You can buy (miniature) 10 year treasury bonds.  Each bond pays guaranteed fixed interest payments of $60 per year, followed by a large principal repayment, $1,000, in 10 years. Importantly, you cannot sell one of these miniature bonds after you have bought it.  You must hold it to maturity.  The money that is figuratively “contained” inside it will be locked up, unable to be used by you in any way, until then.

(3) Equity: You can buy diversified shares of equity, say, the 500 companies of the S&P 500.  The share pays $40 per year in dividends, the dividends grow anywhere from 1% to 4% per year, plus inflation, with a sharp recessionary drop every several years that is eventually fully recovered.  Crucially, the shares have no maturity.  You will never get the original principal back–what you will get back is an ever-growing stream of dividends, which over the very long-haul, will exceed what you put in by many orders of magnitude.

We have not yet stated the per-share price of the bond and the equity.  The goal here is for you to seriously think about the options, as if they were presented to you right now, and identify the maximum price that you would be willing to pay for each share, the maximum amount of cash that you would be willing to permanently exchange for them–which, for you, is the “intrinsic value.”

As a rational agent, what do you need to know in order to determine the “intrinsic value” of each type of security?  For starters, you need to know, or estimate, the concrete specifics of the payout stream. How much is the bond going to pay?  How much is the cash going to pay?  How much is the equity going to pay?

For the bond, you already know the entirety of the future stream–$60 per year, plus $1,000 in ten years. The stream carries no uncertainty in its payouts.  But knowing that alone is not enough.  You also need to know the nominal interest rate that cash in the bank will pay you over the next ten years.  It will not make rational sense for you to pay a price for the bond that implies a return that is any lower than that, any lower than what you can get in cash, because cash also affords you the precious ability to have the money and use it, which the bond does not afford  you.  Therefore the bond needs to be priced to pay you more.

Now, we know that the central bank is going to set the cash interest rate so as to control inflation.  So the true variable that matters here is the future neutral nominal interest rate, the nominal rate of interest that the central bank will have to set going forward, given the structural dynamics of the economy, in order to keep inflation and any other target that it might have–employment, foreign exchange control, financial stability–on target.

In truth, of course, you’re going to demand even more than the expected neutral nominal rate, you’re going to demand a premium to compensate you for the time value of money, the cost of losing the ability to use your money.  How much you demand will be determined by the amount of value that money in hand has for you in comparison with money promised in the future.

How different, for you, is “money now” from “money later”?  The answer will obviously depend on the myriad of complex psychological and economic factors that define your unique personal situation. How much do you value the comfort and security of having access to your money, the ability to use it if you should want to use it, or need to use it?  How much more valuable is that kind of money to you, in comparison with money that will be locked away for a long period of time, inaccessible to you?  How many things are there in the economy for you to buy right now that might tangibly increase your happiness, or the happiness of those you care about?  How old are you, and to what extent is the money needed to fund your desired consumption expenses?  If the money is needed, will the coupon or dividend payments that will accrue on it if it is permanently locked away in a bond or an equity be large enough to fund those expenses?  If so, then you may be fine with seeing it locked away, given that you can get by on the infinite payouts that will accrue.  What are your expectations with respect to inflation?  Inflation eats away at the future purchasing power of money.  High inflation therefore widens the difference between “money now” and “money later”, given that it makes “money later” into “less money.”  All of these variables, and a number of others, will factor into your estimation of the “time value” that money has for you.

To summarize the bond case, then, we’ve identified two variables that matter to the intrinsic value of a long-term treasury bond: (1) the expected neutral nominal interest rate on cash over the life of the bond, which sets the minimum floor for what you can rationally accept from the bond, given that you have the alternative of holding cash, and (2) the time value of money, which you ultimately have to specify for yourself, given the unique psychological and financial details that characterize your individual situation.

For the equity case, the evaluation is more complex.  We need to estimate the future growth of the dividends, and by extension, the future growth of the earnings out of which they will be paid (and which will pay for their future growth).  In the scenario, we set a range of 1% to 4% after inflation, but that’s a huge range–any information that pushes the number in either direction is going to be very important.

We can separate the growth of dividends into two components: real per-share growth, and inflation.  The first component is determined primarily by the health and dynamism of the underlying economy, and by the efficiency and capital allocation skill of the aggregate corporate sector. The second component is driven by culture, demographics, supply constraints and policy.

The two factors that were relevant to the intrinsic value of the bond–the expected neutral nominal interest rate and the time value of money–are just as important to the intrinsic value of the equity.  As with the return produced by holding the bond, the return produced by holding the equity competes directly with the alternative of holding cash in the bank and collecting the future neutral nominal interest rate.  Similarly, holding the equity instead of the cash entails loss of a large amount of money that would otherwise be accessible.

What we end up with, then, are four variables that determine the “intrinsic value” of the equity: (1) the expected neutral nominal interest rate, (2) the time value of money, (3) the expected future rate of real per-share growth, and (4) the expected future rate of inflation.

Now, here comes a critical move.  We can combine (4) and (1) into a single variable, the expected neutral real interest rate.  Going forward, what real interest rate, after inflation, will the central bank have to set in order to maintain inflation, and any other targeted macroeconomic variable, on target?  That rate is critical, because it expresses the difference between (a) inflation, a crucial component of the nominal growth that the equity payouts will exhibit, and (b) the nominal interest rate that the cash holdings will earn.

The Fed Model

The Fed Model is a popular a method of measuring equity valuations.  The model assesses valuation by comparing the earnings yield on equities to the long-term government bond yield. When equity earnings yields are substantially higher than the long-term government bond yield, equities are said to be cheap.  When equity earnings yields are not appreciably higher than the long-term government bond yield, equities are said to be expensive, or at least neutrally priced.

In practice, the Fed Model has caused a number of analysts to push back on the growing consensus that the US stock market is expensive, while Emerging Market stock markets are cheap.  These analysts acknowledge that earnings yields in the US are lower than in the Emerging Markets (or alternatively, that P/E ratios in the US are higher than in the Emerging Markets), but they point out that we cannot talk about yields and P/E ratios in a vacuum.  We have to compare them to the available alternatives, the attractiveness of which are captured by prevailing interest rates.

But this way of thinking is partially wrong.  It ignores the fact that interest rates are typically set at low or high levels in response to low or high levels of another variable that matters greatly to equity returns–inflation.  Why has the US  central bank set the interest rate at a low level? Because the US does not have enough inflation. Why has the Brazilian central bank set the interest rate at a high level? Because Brazil has too much inflation. The low inflation in the US contributes to an environment of low nominal earnings and dividend growth, and therefore low nominal total returns, all else equal (and note that all else is not equal, in this case).  The high inflation in Brazil (or Argentina or Zimbabwe) contributes to an environment of high nominal earnings and dividend growth, and therefore high nominal returns, all else equal.  The Fed Model fails to capture and factor in the impact of this crucial difference.

If we’re going to connect P/E ratios to interest rates, as the Fed Model tries to do, the interest rates that we should use are real interest rates, interest rates that take out expected future inflation, which is a significant component of nominal equity returns. When we do this, we see that a number of emerging markets with high interest rates and high P/E ratios, such as India, deserve to have high P/E ratios, because their real interest rates are very low, if not outright negative (making cash and bonds that much less attractive in comparison with inflation-linked equities).  Similarly, a number of countries with low interest rates, such as Japan under pre-Abenomics deflation, deserve to have low P/E ratios, because their real interest rates are high (making cash and bonds that much more attractive relative to inflation-linked equities).

Foreign Equity Investing

This dynamic extends quite elegantly to the realm of foreign equity investing. To use the example of Brazil, Brazilian equities currently sell at very low P/E multiples–at last check, around 8-9 times, with correspondingly high dividend yields and substantial room for P/E multiple expansion over the long-term.  For this reason, many US investors, frustrated with the lack of attractive options at home, have explored the country as a potential investment opportunity.

Suppose that you are a US investor that wants to capture the return potential of the Brazilian equity market.  But you want to capture that return in dollar terms–the terms of your own currency.  If the Brazilian market goes up 200% over the next 10 years, you want the value of your Brazilian investment, in your own currency, the Dollar, to achieve that same return.  The only way that you can make this happen is by hedging the currency.  You would go simultaneously long the Brazilian stock market, and short the Brazilian currency, the Real.  Then, your return in dollars would exactly mimic the local currency return of the Brazilian stock market.

But there’s a problem. The cost of shorting the Brazilian currency is the Brazilian interest rate; you will have to pay that interest rate to whomever you borrow the currency from in your short.  Right now, the rate is quite high, north of 10%.  That 10% will represent a significant drag on your returns.  For this very reason, it’s impossible for you to create a dollar-denominated investment that will exactly track with the Brazilian stock market. The best you can hope to do is create an investment that tracks with the Brazilian stock market minus 10% per year.

Not all is lost, of course.  Your investment might still produce an attractive return, even in the presence of the high carry.  The interest rate in Brazil is high, 10%, but that’s because inflation in Brazil is very high–well north of 6%.  The 6% inflation is going to add to the nominal growth in Brazilian earnings and dividends.  When combined with the high dividend yield, and the significant multiple expansion that is likely to occur as sentiment improves, the return that the investment might be able to make up for the 10% carrying cost.

What we need to do in a valuation analysis is combine these two numbers–the inflation and the interest rate–since they offset each other in terms of their effects on the return. The inflation adds to the return, and the interest rate–which is the carrying cost–subtracts from it.  The combination of the two,  of course, just is the real interest rate, which, you will recall, is what we found to also be a critically important variable in the determination of the intrinsic value of domestic equities.

The real interest rate in Brazil is 10% minus 6% which equals 4%–on the high side globally. For this reason, Brazil probably should have a lower P/E multiple than the developed world, where zero or negative real interest rates have become the policy norm.

Posted in Uncategorized | Comments Off on Intrinsic Value: Interest Rates, Inflation, and the Forgotten Concept of the Time Value of Money