Monday, March 9, 2015

The huge potential value of NGDP Level Targeting, a CAPM and Risk Trading perspective

If the capital asset pricing model is valid at all, then the potential gains from NGDP targeting are enormous.

First, let's consider the capital asset pricing model, which describes equilibrium returns to securities thusly:
Basically, there is a risk-free expected rate of return.  To earn more than that with a diversified portfolio, you must take exposure to risky assets.  Note, though, that since idiosyncratic risk can be reduced through diversification, it ceases to be a risk in a diversified portfolio.  The risk that remains is market risk - risk that cannot be diversified because it correlates with volatility in the full market of securities.

Market beta is a product of two unmeasurable factors.  (1) The expected amount of volatility in equity returns (in perpetuity) and (2) the aversion that the marginal investor has regarding exposure to that volatility.  Since we can roughly measure Beta, but the two factors composing Beta can't be measured, and are conceptually vague, even while being intuitively obvious, CAPM, like EMH and so many other financial concepts is not falsifiable.  That doesn't mean it doesn't reflect some basic and undeniable truths.

One way to think about it is to compare earnings from risk free bonds to equities (or other risky securities):
For a risk free bond, y will fluctuate over time, so it is unknown at the point of purchase.  The risk of changes in y increases with the maturity length (the duration) of the bond.  But, x is a constant.  So, short term T-bills are truly risk free, and their required return is the base, risk-free return for delaying consumption.  Bonds with longer maturities fetch a premium because changing y adds risk which affects the entire market basket of securities.

The Equity Risk Premium [E(Rm)] above, is the price for accepting an additional unknown variable - an uncertain x.

Think about the power of this idea, with regard to NGDP level targeting.  All risk is described as market volatility.  And NGDPLT, as conceived, practically eliminates market volatility (at least in terms of total earnings).  Here is a chart of the share of GDI going to capital (minus corporate taxes and income of homeowners).  This has ranged from about 16% to about 20% in the modern era.  But, note that nearly all of the volatility comes from NGDP shocks.  If we remove the cyclical shocks, the level of operating profits (interest plus net profit) is very stable.  In a stylized version of NGDPLT, there is no market risk.  Equity earns 5% more each year, because we determine that NGDP will grow through monetary policy, and proportional capital income, in equilibrium, will change very slowly.  So, whatever aversion to risk there is, volatility is nil, and E(Rm) = Rf.  The required return on risky assets = the risk free rate.

For the sake of discussion, let's limit this thought experiment to the domestic economy within the NGDPLT regime.  Even if NGDPLT doesn't achieve perfect earnings stability (for the diversified portfolio), it would clearly move significantly in that direction.  Here is a chart of 2 year changes in GDP and in Enterprise Value of Nonfinancial Corporations.  A simple regression, by quarter, of these measures accounts for about 1/4 of changes in Enterprise Value.  And, this is simply the real time change in GDP.

I have seen some reservations about NGDPLT because of the idea that equities exhibit unusual volatility, which is attributed to some sort of behavioral bias, fickleness, or serial correlation.  But, in statistical terms, the jury is still out about whether markets reliably exhibit first or second order mean reversion.  And, in practical terms, many former economies, which certainly didn't revert to the mean, lie in the trash bins of history.  The notion that market values should not react to negative GDP shocks with lower growth expectations or higher equity risk premiums seems na├»ve to me.  And, conceptually, the volatility in equity values that coincides with GDP shocks must be related to expected future GDP.  NGDPLT eliminates that problem.  There is no significant uncertainty about future GDP under NGDPLT.  Even the volatility in equity values that can't be empirically attributed to GDP volatility, must be conceptually attributed to it.

So, what happens in an economy with a small Equity Risk Premium?  First, I find it useful to tweak the concept of CAPM a little bit.  Over time, I find that total real expected returns to unlevered corporate capital [Rf + E(Rm)] are fairly stable - reverting to around 8% over time for all nonfinancial corporations - maybe a bit less for S&P500 firms because of the lower liquidity premium.  (Most fluctuations in equity values can be explained by changing earnings and changing growth expectations, not a wildly fluctuating level of required total returns.)  So, instead of thinking of returns as additive, I think it might be more helpful to note that there is a fairly stable return to productive capital, and capital owners are trading risks when they share those returns.  So, (in real terms) if Rf is 3% and E(Rm) is 5%, it might be better to think of it in terms of there being a 7% to 8% real return to unlevered corporate assets, and debt holders are willing to accept a 5% discount in order to eliminate cash flow volatility.  (Note, when considering CAPM in terms of the equity premium, I use long term treasury bond rates as Rf, in order to match durations.  If we use the short term rate, we end up creating all sorts of confusion by conflating maturity premiums with equity premiums.  A series on subtle conceptual issues with CAPM and beta has been collecting dust on my to-do list.)

In the NGDPLT world, there is minimal cash flow volatility (for the diversified market portfolio), so there is no reason for a deep discount.  In that case (again in real terms), we would expect to see a high Rf (say, 6%) and a low E(Rm) (say, 2%).  What effect will this have on capital allocation?  This means that the premium required for a high risk investment is very low.  A zero beta investment would require a 6% real return, while a 2 beta investment would only require a 10% real return.  In today's environment (Rf = 1% and E(Rm) = 6%), a zero beta project would require a 1% real return, compared to a 13% real return for a two beta project.

In the low interest rate environment, there is tremendous pressure to invest in low productivity projects.  In the high interest rate environment, there is much more incentive to push capital into transformative, disruptive, risky ventures that will create, and gain from, high GDP growth.  And, this is exactly what we see in the modern era.  Low E(Rm), in the 1960s and 1990s, were associated with high growth.  (Note, they were also associated with high compensation levels.  Also, note that low E(Rm) came during periods where the business cycle had been tame.)  Also, the 1960s and 1990s were associated with relative declines in real estate investment and valuations.  Capital was being pulled, instead, into industrial and commercial production.  Contrast this with the high E(Rm) 1970's and 2000's, where real growth declined, compensation sagged, and capital poured into real estate instead of commercial ventures.

Additionally, volatility of "y" would also decline, so the maturity premium would be lower.  This would induce more investments into long term projects.  And, fixed incomes and savers would benefit from high real risk free returns.

As I mentioned, during periods with low risk aversion in the late 1960's and late 1990's, total returns to productive capital have remained fairly stable.  Most of the shift in risk premiums caused real Rf to move up, especially on the short end.  So there was little maturity premium and little equity risk premium.  So, we already have evidence of E(Rm) moving below 3% because of minimized NGDP volatility.  These periods might have been associated with a slight reduction in required total capital returns, but even the extreme equity gains of the late 1990's were mostly due to high growth expectations.  Real Rf (based on 10 year treasuries) was between 4% and 5% for most of the 1990's, before any equity premium would even be added.

In this last chart, the dark red line is the average annual deviation from 7% NGDP growth over the previous 5 years and the light red line is the standard deviation in NGDP growth over the previous 5 years.  The light blue line is my estimated unlevered equity risk premium from the Fed's Flow of Funds report, and the dark blue line is Damodaran's equity risk premium for the S&P 500.

So, I wouldn't expect to see a huge effect on required returns to equity as a result of this regime shift, and I'm not sure we would see much of a shift in valuations in general, since NGDP growth expectations would be moderate, by definition (edit: although, if all these factors lead to higher RGDP growth expectations, valuations would rise).  Instead, I would expect to see a huge shift in the types of investments that corporations make.  Maybe we are too deterministic about past economic development.  Would the technology boom have been as overwhelming in an environment with a 5% equity risk premium.  Think of the amount of capital that was being invested into high risk ventures that depended on extreme economic and cultural developments and didn't promise payouts for years.  Could those investments have been made in a 5% ERP context?  I don't want to be hyperbolic, but I suspect that we underestimate this factor.

And, considering we have already seen E(Rm) fall into the 2%-3% range during these periods that were shaped by just a few years of lowered NGDP volatility, it doesn't seem outrageous to me to think that NGDPLT could be associated with 1%-2%.

NGDPLT could be absolutely transformative - higher growth, higher compensation, more risk taking, even though absolute returns to risk would be low.


  1. I am looking for a leader who wants to focus the next part of their life on figuring out how to make NGDPLT happen. Please contact me with any leads.

    Kenneth Duda
    Menlo Park, CA

  2. Kenneth,
    It could be that NGDPLT - at least in the U.S. - may have to be "proven" through applied means, that is by groups who are willing to organize their energies internally so that production and services can be realized locally. That would mean an internal transmission which makes it possible for economists to observe the potential of income with resources in local environment.

  3. Hi Becky, I did not understand what you wrote. Would this "group who is willing to organize their energies internally" have their own currency that floats against the USD?


  4. Ken,
    You are familiar with charter city concepts. I would like to see versions of these in which citizens are free to opt out of all government transfers to take part in local investment structures. Services would be generated through coordinated time, in which education becomes part of work flows for local populations. This coordinated time base serves as initial income, by which to organize local investment options to augment both one's time use and the further production options communities choose. Otherwise these communities would remain a part of the economies (state and national) around them in terms of tradable goods manufacture and monetary flows. The internal focus is to measure non tradable assets and services (regular records for the Fed) to bring them back into balance with traditional production. Coordinated services, because there is no residual time or monetary debt, would be newly generated time based wealth. The assets and services (non tradable) combination is the monetary transmission part that would be separate from surrounding economies. This structure could also show how income and time aggregates could better correspond with resource use than is now the case. A central goal in all this is to provide a reminder that time aggregates are central to economic activity, as well.

    1. Well, I really haven't thought much at all about charter cities, but I will say that there is no way that a charter city could practice NGDPLT unless it also had its own independent currency (i.e., floating against the world's major currencies) as well as its own banking system and central bank. It is unlikely that a city is an optimal currency area, and I'm finding it a little hard to predict how the benefits I imagine for NGDPLT would trade off against the overhead of trying to manage a currency / banking system.

      It sure would be nice if there was a cost-effective way to run some experiments !

      Thanks for the thoughts.


  5. Excellent article Kevin I think you've captured an important "side-effect" of NGDPLT. I wonder what's the best marker to measure. Many people say that GDP and inflation are almost meaningless and I think they have a point. I find myself thinking more of something like the median wage and the nominal growth thereof annually. What is the output gap exactly and can we measure it accurately. And what are the best measures to take to raise NGDP.
    Keep it up.
    PS - I saw your comments on the money illusion - why don't you weigh in with your preferred tax code.- on here I mean.
    PPS - the xls for housing valuation was brilliant - I've copied it and will use it next time I buy or sell a home(s) - I hope that's OK!?

    1. Thanks, Sean. As I work through the housing series, I will probably have more to say about real estate tax issues. Otherwise, I generally would prefer either a consumption tax or a very simple income tax coupled with a guaranteed income, and no taxes on capital. But, I'm not sure that I have that much to say on the topic, and I doubt that my favored policies are practically or politically reasonable.

      I realize that it may not be helpful from a public policy standpoint, but I think my competitive advantage is in finding things on the margin that are off kilter in some way, and considering the ramifications. I'm not sure that my blog would add value in terms of broad, sweeping policy ideas.

      On your NGDPLT points, it seems as though the IRS should have pretty detailed information coming out of payroll tax data that could be tracked, like Case-Shiller tracks home prices, with weekly or monthly real-time data on changes in individual wage levels. It seems like any number of proxies for production could work.

      I'm glad you like the model. Eventually, I expect to tweak it with some of the BEA data I've been playing with. In the end, I think I will frame the homeownership question in terms of the benefits of ownership (eliminating principal-agent problems) and the cost of non-liquidity & non-diversification in holding such large exposure to a single asset. I think there may be interesting implications from that framing, in terms of aggregate economic gains and losses. Hypothetically, I suspect that we would gain much more value if homeownership was framed like a call option instead of like a highly leveraged purchase. But, I have a lot still to work through on that series.
      And, please feel free to use the model. You can eventually repay me by letting me stay every now and then in your (eventual) portfolio of luxury villas. We can share some wine on the veranda.

  6. TravisV here.

    Any idea why 5-year inflation expectations (breakevens) have surged so rapidly lately??

  7. TravisV here.

    Hmm, Scott Sumner's explanation: "Travis, Probably because the recent fall in the CPI is behind us now, whereas 2 months ago people knew it was coming, but it wasn’t yet in the data."

  8. TravisV here.

    Kevin, here is why I love this blog post. I interpret you as reading this:

    "My personal view is that when the ERP is above 4%, the stock market offers an adequate risk premium, and when it is below 4% it is less compelling and more risky. The all-time low was set in 1999 at 2.0%."

    and saying "Yes, I agree with Christopher Mahoney, except I think he's too conservative! In the medium-term, the Equity Risk Premium could easily fall as low as 3.0% and might even fall below 2.0% before it's all said and done!"

    Why? Because over time, we should expect central banks to gradually get better and better and better at delivering stability.

    Is that a reasonable interpretation (although perhaps a bit overstated)?

  9. That's pretty much what I'd say. If anything its an understatement.
    My nitpick with the article would be that his causation is wrong. Falling earnings cause the value to drop. This causes bond rates to fall, and ERP to rise, but I don't think this leads to considerable change in the total equity discount rate. This leads to his poor conclusion that equities will correct when rates rise, which is not true.
    There is a lot going on there. For instance, falling earnings probably affect high beta equities more, but the changing ERP would also change relative valuations across beta. I'm not sure most financial research treats all of this with enough care. I think a lot of the critiques of CAPM are failing to specify and track all of these moving parts. Eventually I will do a long series on that.

    I don't know what's happening on inflation. I expect to eventually take a strong position, but I'm on the sidelines now. It depends on fed discretion. I expect this could be volatile for a little while. The position I might take in the current context would be a neutral position that was short implied volatility and long actual volatility.

    1. TravisV here.

      Thanks Kevin!

      I agree with your critique of Mahoney. I think you might agree with me that there's an "optimal" rate of NGDP growth of perhaps 5%-6% that would maximize stock valuation multiples.

      I still think Mahoney deserves credit for his overall analysis, however. I think he's closer to reality than 95%+ of equity analysts out there. I thought this article by Mahoney was also brilliant:

      P.S.: Check out the last paragraph of this post:

      While I share this guy's intuition against growth stocks, I prefer Colgate and Nestle to AT&T, IBM and GE because the economic moats of Colgate and Nestle are stronger. Also, Colgate and Nestle are less capital intensive (require less CapEx reinvestment) than AT&T and GE.

    2. Hmm. I didn't like the first article. He needs to be more careful about real vs nominal, especially if he's playing around with premiums as ratios vs differences. I think it creates some confusion for him. Also, volatility is not a refutation of EMH. Quite the opposite. Compare the wheat market to the market for legal representation.

      I think you're right that the end of the Joe Leider post is practical evidence of what I'm talking about. And even regarding broad market valuations, higher equity values wouldn't come from the lower ERP, since interest rates would generally be high, to total required returns would still be about the same. Higher equity values would come from higher growth expectations.

      Although, while this describes the 1990s, it looks like the low ERP period of the late 60s might have seen more moderate growth expectations and lower total required returns. It's hard to say because it depends on the inflation adjustment, and inflation was starting to rise, so I don't have a precise measure for the period.

    3. TravisV here.

      Kevin, you wrote:

      "regarding broad market valuations, higher equity values wouldn't come from the lower ERP, since interest rates would generally be high, to total required returns would still be about the same. Higher equity values would come from higher growth expectations."

      It's not quite clear to me where we disagree. Here's my simple way of thinking about it.

      The ERP is a proxy for "future probability of a massive recession." If the odds of a massive recession increase, then real growth expectations decrease and vice versa.

      So when the ERP decreases, it's not merely signaling that (1) future cash flows will be more stable. It's also signaling (2) a higher real growth rate for future cash flows.

      Sure, theoretically, it could be 100% (1) or 100% (2), but that's extremely unlikely. And it might be impossible to disentangle (1) from (2).

      So forecasting a lower ERP is effectively the same thing as forecasting a higher real growth rate.

      That sentence might sound simplistic, but as far as I can tell, it's still correct.

    4. Good point Travis. Especially from a security valuation standpoint, these definitely get tangled up.

      But I do think there are conceptually different things going on - basically the difference between volatility and growth. So, we might think of an economy with a long term real growth rate of 3% and a standard deviation in growth of 2%, with highly negative skew. That situation definitely is an input in ERP. So, first, you're right that 3% growth with less than 1% should lower ERP. I think this mostly is mitigated by rising Rf, in terms of corporate equity valuations. But, if we have some permanent losses from economic downturns, maybe average growth goes to 4% when we minimize them. I agree this is reasonable. This would lead to higher valuations because of higher growth.

      But, I think there would be a separate, second order effect due to the lower hurdle rate for required returns on high risk projects. In effect, market beta, while definitionally stuck at 1, would represent more disruptive capital investments. There would be more idiosyncratic risk, even while market risk remained very low. This would raise real growth even further. The risk of owning the market will be low, but the risk of owning an individual stock will be higher.

      On this the Austrian liquidationists have it wrong. They say recessions help wring out inefficient producers. But, having more moderate business cycles would do an even better job of this - and it would happen in the trenches, where it should happen, instead of being random and nationwide. There would be more Amazons replacing Borders and fewer workers laid off for 3 years just to be hired back again to do the same basic job.

      Thanks for the comment. It made me clarify my thinking a little more on this.

  10. Kevin,

    Why not specify the Chen, Roll, and Ross (1986) APT model, which explicitly includes industrial production? You could put NGDP right into the asset pricing equation if you wanted.

    1. You might have something there. APT just gets conceptually messy for me, so I'm more comfortable in a CAPM framework.

  11. This entire post is tragically flawed. Folks, examine your priors! What are you saying when you assume "There is no significant uncertainty about future GDP under NGDPLT"? You are assuming that the quantity theory of money always holds in such a way that velocity is constant. But velocity is not constant. You guys are bent on reshaping the world in your own image and likeness, but the entire exercise is comically absurd. And Duda, you have too much time and money on your hands.