Milan Seminar: November 19 — With Anatole Kaletsky & Cedric Gemehl

Seminars Close
Gavekal Research

The Savings Glut’s Long Life And Slow Death

Demographics, some say, can explain two-thirds of everything (the problem of course is knowing which two-thirds). Demography has certainly  influenced  long term asset-price trends. For roughly 35 years, from the early 1980s until just recently, global capital markets were lent wings by a major demographic tailwind. Changes in the age structure of the world’s population created a “global saving glut”, to use the term  popularized in 2005 by then Federal Reserve governor Ben Bernanke. The high volume of savings looking for a home bid up asset prices massively, propelling secular bull markets in both equities and bonds. It was a great time to be in finance.

Now that demographic tailwind is fading. And a decade hence it is likely to reverse, turning into a headwind for asset prices. To chart the impending changes, this paper introduces a new measure, the Capital Providers Ratio, which translates demographic shifts into trends in real interest rates, and so to asset prices. The basic conclusion is that real rates will stay at roughly their present low level for another 10 years or so, and then start to move up as the supply of excess saving begins to thin out.

Saving / investment stage 1: The Baby Boom Effect

Globally, saving must equal investment. But capital is mobile, so it does not follow that savings must equal investment within individual economies. Even China, which restricts private capital flows, still sees large capital outflows through official—and unofficial—channels.

As a result, different countries show different propensities for saving versus investment. Analyzing data from 85 countries between 1960 and 2005, economists at the Brookings Institution found that a country’s propensity to save is likely to exceed its propensity to invest when the population is skewed toward the ages of 35-64.

This pattern of saving and investment makes intuitive sense. Children earn nothing and consume plenty, so they are a drain on national saving. At the same time they call for investments in schools, clinics, larger homes, and so on. Higher education extends the period in which young adults are saving little, while much is invested in their education.

After they enter the workforce, most young adults will not earn enough to do much saving. And these young workers either invest a lot themselves (buying their first home and other durable goods) or encourage companies to invest on their behalf (directly, for example through worker training, or indirectly, e.g. by building factories for them to work in). So when a country has a baby boom and has a population increasingly skewed toward children and young adults (0-34 years) it will show a stronger propensity to invest and a relatively weak propensity to save.

In a single country, this shift might manifest itself through more borrowing from abroad. For the global economy, a closed system, saving and investment must be brought into balance by a change in the clearing price for saving—i.e. a higher real rate of interest. This is what happened in the West between 1946 and 1980, as the Baby Boomers were born and grew up.

Stage 2: The Middle-Aged Worker Era

As the population ages and the dominant group becomes older workers (aged 35-64 years), the situation reverses. These are the peak earning years for most people, so they can save a lot. At the same time, their investment needs fall; they have finished their education and have already bought their homes. Companies, facing both an aging customer base and a workforce nearing retirement, will be more reluctant to make new long term investments.

In this era, the propensity to save exceeds the propensity to invest. Individual countries in this situation will tend to export more capital. And if it happens on a global scale, again the clearing price must adjust. The real rate of interest must fall to the point at which savings and investment balance. This roughly describes the period from 1980 to the present: first as the western Boomers matured into their peak-earning years, and then as demographic trends in China and other emerging economies generated plenty of excess savings.

The largest demographic movement in the developing world was China’s baby boom in the late 1960s and early 1970s, before family planning measures that culminated in the one-child policy brought down birth rates. When this group started to have their own children, they created an “echo” baby boom in the early 1990s—a cohort which will reach prime earnings age in about 7-10 years, tilting the savings-investment balance a shade more towards savings.

Stage 3: The Retirement Home

The final stage of the long demographic arc comes when the baby-boom generation starts aging into retirement (assumed to be around age 65). Now the propensity to save falls again, because the typical retiree is a net drain on national savings. Once they stop working, the elderly start to consume their own savings and/or draw on national savings via a public pension plan or national healthcare system. Meanwhile, at this stage investment demand makes a modest rebound, probably because of the need to build new hospitals or nursing homes.

As a result, a very old population, like a young population, sees its propensity to save fall below its propensity to invest. Capital must be borrowed from other countries or, if it is a global phenomenon, the real rate of interest must rise to bring savings into equilibrium with investment. This is the situation the world will find itself facing from roughly 2027 onward.

To make it useful, apply CPR

My Capital Providers Ratio (CPR) presents these shifts in saving and investment behavior in a way that helps explain the demographic impact on interest rates and asset prices. The basic idea is that countries with a relatively large cohort of people aged 35-64 will have excess savings and will tend to export capital, putting downward pressure on the global real interest rate. Countries with relatively large populations younger than 35 or older than 64 will tend to dis-save, and push up the global real rate.

So to calculate the CPR, we divide the number of people in the 35-64 cohort by the sum of those in younger and older age cohorts. Looking first at the national level, there are no surprises about the identities of the big capital exporters. Japan has been a consistent large-scale capital exporter for 40 years, and was best-in-class in the 1980s. But with population aging and its CPR declining towards the global average, Japan no longer features as such a big capital exporter. Today and for the next 10 years, the biggest capital providers will be China and Germany. Smaller but still sizable contributions will come from South Korea and Russia, which have high CPRs but much smaller economies.

Germany will decline as a capital provider after 2030, but China is likely to ramp up quite a bit over the next 10 years, and along with Russia and South Korea will remain an important net provider of capital until 2040. After that, aging will take its toll and these countries’ propensity to save and export capital will decline. Brazil and India will then take up the baton in the excess-savings relay.

The country-level data is interesting (and I will explore it more in forthcoming reports). But what we really want to know is the global saving-investment balance and its impact on interest rates. And a moment’s reflection should make clear that, at least under present conditions, the relationship between demographically-driven savings balances and interest rates should be more stable at the global level than at the country level.  

The reason is simple: in a world with low barriers to capital mobility, global effects can overwhelm parochial ones. Generally speaking, when the CPR rises—meaning that saving rises faster than investment—real interest rates should fall (and equity multiples rise). If all that mattered was a country’s own saving/investment balance, Japanese rates should have bottomed in the early 1990s when its CPR plateaued, and Germany should have seen its rates bottom in 2005. One reason they did not is that the global saving surplus continued to rise until at least 2010.

To calculate the global CPR we need to make a choice: weight countries by population or by GDP. On a population-weighted basis (using United Nations population projections), the global CPR is still rising and will not level off until 2030 or so.

But a nation’s ability to export capital is determined by the size of its economy, not its population. So a GDP-weighted global CPR measure will be more accurate.  The red line in the chart below shows an average of the CPRs of the 32 largest economies in the world today weighted by PricewaterhouseCoopers’ long-term GDP projections. These countries make up about 85% of world GDP, and so serve as a decent proxy for the world economy.

On this basis, the world CPR leveled off around 2012. Over the next 10 years it will edge slightly higher once again, thanks largely to the savings contribution of China. But from roughly 2027 onwards it will decline. This implies that global real interest rates are now close to their structural bottom, and may start to rise structurally after 2027.

For reference, I have left in the more optimistic population-weighted measure (blue line).  It enables us to imagine the trajectory of global CPR if per-capita GDP in high-population, low-income countries such as Nigeria, Egypt and Indonesia grows faster than in PwC’s projections. On the whole, though, the GDP-weighted series is not overly skewed in favor of today’s winners, since it already includes the three countries just mentioned plus India, Vietnam, the Philippines and South Africa—all of which have big populations, strong demographic outlooks and are projected by PwC to enjoy rapid GDP growth in the coming decades.

Watch out for falling long bonds

In short, global real rates have ended a three-decade period of structural decline, will continue to bounce along the bottom for another 10 years or so, and will then begin an inexorable rise. The next question is: What does this mean for asset prices?

The first observation is that since the early 1980s, investors could have reaped rich rewards simply by holding a balanced portfolio of 50% equities and 50% extremely long duration bonds, such as 20-30 year zero-coupon US treasuries. This has long been Charles’s preferred allocation (see Portfolio Building In The Time Of Disinflation).

Why did this strategy work? As interest rates went into structural decline between 1982 and the present, equities and bond prices enjoyed a secular bull market—together. But while the long-run trends of the two asset classes were aligned, their short-run variations tended to show an inverse correlation, especially at times of market stress. Thanks to this combination of long-run positive correlation and short-run negative  correlation, over the last 35 years a balanced portfolio has offered similar long-term returns as an all-equity portfolio, but with lower volatility.

What worked extraordinarily well during Charles’s career may not work so well over the course of my career. Charles’s approach was to use a very long treasury bond, for example a 30-year zero coupon US treasury—as the basis of his bond holdings, only reducing duration tactically when bonds looked overvalued or overbought.

This strategy could potentially have another 10 years of life in it. But if my demographic analysis is correct, then buying a 30-year bond today may eventually turn out to be dangerous—because it would lock in today’s low interest rates beyond the 2027 structural turning point implied by my Capital Providers Ratio.


Interactive chart

Forward projections of current demographic trends suggest the new default for the bond portion of a balanced portfolio should be notes maturing within the next 10 years, before 2027. Investors can tactically reduce duration when valuations warrant (and they should consider doing so today; see QT And The Treasury Quandary). They can also extend duration tactically when long bonds offer a sufficiently high yield. But investors should be careful about buying bonds with maturities beyond 2027, making sure that the term premium sufficiently compensates them for the fact that the demographic winds are likely to move against existing bond-holders in roughly 10 years time.

Demographic analysis may also help solve a longstanding equity market puzzle: what lies behind the structural increase since the early 1980s in the S&P 500’s equity multiples? Anatole has long pointed out that investing based on a return-to-the mean for P/E ratios has been a money-losing strategy for the past 25 years. In all that time, it has issued just one buy signal—in early 2009. As the next chart suggests, the rising world CPR since 1980 may have driven up structural equity valuations.

Conclusion

This paper is just a preliminary study of the slow-moving demographic forces that drive long-run trends in asset prices. More detailed work remains to be done. But it does suggest answers to two questions that are very much at the center of investor concerns today:

1) Will interest rates remain low forever? Or is the bond bubble about to burst? The answer here is “no and no.” Globally, the prime-earning-age cohort that produced the great savings glut of recent decades is no longer growing significantly. But it will not start to shrink until roughly 2027. In other words, savings will remain plentiful and—cyclical considerations aside—interest rates could remain low for another 10 years. After that, however, demographic forces will probably work to push real interest rates higher.

2) Can the secular rise in equity valuations over the last 35 years continue indefinitely? The conclusion is similar. Since the early 1980s, global demographic trends have led to a glut in savings which has propelled a long term expansion in equity multiples. With the global Capital Providers Ratio now leveling off, it is likely that equity valuations will range sideways over the next 10 years or so. After that they could enter a long term contraction.

22 Comments

  • >Anonymous

    Anonymous

    17/10/17

    Great piece thank you. Any general thoughts on how this plays out if people start living until they are 125?

    Reply
    • >Will.Denyer

      Will.Denyer

      20/10/17

      Dear Reader,

      Good question. If people start living to 125, it is hard to say whether this would be negative or positive for real rates, or have little net effect at all. Consider the following possibilities: 

      Longer life could put UPWARD pressure on real rates if:

      • it results in longer retirement periods. This implies a larger/longer drain on national savings, and thus would be a force for higher real rates.
      • If life is extended as a result of the economy commiting more resources to healthcare and other life-extending measures, then this too could be thought of either as a detractor from savings (if you consider it consumption) or an increase in demand for investment. Either way, it implies higher real rates. 

      Alternatively, it could put DOWNWARD pressure on real rates if: 

      • the working period of life is extended (more so than the retirement period). This would add to national savings and push down real rates.
      • And if life is extended thanks to technological breakthroughs (rather than just commiting more resources to the problem) then this need not put upward pressure on real rates

      It could also be that real rates are not be affected much at all, if the additional saving allowed by an extended working period more or less balances out the demands of an extended retirement period, and if most of the life-extension comes from technogical advances rather than commiting more resources to the problem. 

      I wish I knew which was more likely, but I don't at this point. If you have any thoughts on the matter please share.

      Thanks for the kind words and good question.

      Will

      Reply
  • >Anonymous

    Anonymous

    17/10/17

    Will,

    Thank you for this piece. It's great.

    How much would your conclusions vary if GDP weights change over time? Demographics are highly predictable but GDP mcuh less so.

    Best,

    Reply
    • >Will.Denyer

      Will.Denyer

      20/10/17

      Thank you, and good question. 

      China is the most important capital provider of the coming decades. As such, I decided to test the sensitivity of my World CPR estimate to changes in the GDP outlook by cutting in half the growth that China is projected to experience between now and 2050, according to PwC.  After doing that, somewhat suprisingly, the story does not change all that much.  The hump that you see in the World GDP-weighted CPR between 2025-2030 (see original chart in the piece above) is dumbed down in the test chart below. However, the general story remains the same--that the CPR is more or less flat until late next decade and then declines thereafter.  

      Changes in GDP outlooks will certainly have an effect, but based on this test it seems it is not hypersensitive--to even fairly major deviations from teh projections (like cutting growth in half for the biggest capital provider). Of course, if China were to collapse or see its GDP (in USD terms) fall significantly relative to the rest of the world, then that would change the picture more dramatically. So would WWIII.

      Good question. 

      Will

      Reply
  • >Anonymous

    Anonymous

    18/10/17

    Will,

    As countries like India approach middle-income status, won't the saving/investment/consumption dynamic change?  At the very least the stock of savings should become more mobile.  Could you be underestimating the length of the savings glut and its impact on financial markets?  A very interesting piece.  Thank you for writing it.

    Jim

    Reply
    • >Will.Denyer

      Will.Denyer

      20/10/17

      Dear Jim,

      Thank you.

      It is my view that savings are already very mobile--including between EM and DM countries. Just look at all of the FX reserves that have been acquired by EM central banks. That is savings from that EM country being lent to the US (it may be forced savings due to currency manipulation, but it is still saving being exported to the US). So I don't think that is going to change.

      Saving/investment/consumption dynamics will indeed evolve, and this is an admitedly crude first past at this analysis.  You, I or someone else could hire a team to do an indepth analysis of what is the approriate way to measure the CPR for individual countries at different points of time, past and projected future. I could be wrong, but my guess is that after all that work we would come up with an only modestly adjusted picture. I don't plan to do this for every country. I will see if I can refine my outlook for the key players however (China, US, India, Germany, Japan...) and if that provides anything new to the analysis I will let you know. For now, I think the most effecient thing to do is recognize that an important demographic change looms. I have estimated that the next turning point is around 2027, but given that dynamics may change I would consider there to be a signficant margin of error (say give or take 5 years?). 

      One last thing... According to PwC's projections, which I use for the GDP-weights, India already rises from being the 7th largest economy (in USD terms) in 2016 to the 3rd largest economy in 2030 and also out to 2050 (it is third after China at first, and the US second; Indonesia comes in a distat 4th by their estimate). So, the measure already factors in a lot of growth from India.  

      Best regards,

      Will 

      Reply
  • >Anonymous

    Anonymous

    18/10/17

    Great piece!

    Reply
  • >Anonymous

    Anonymous

    18/10/17

    Will, informative piece.  What do you think about the argument made in a recent BIS paper (http://www.bis.org/publ/work656.pdf) that inflation will also rise structurally along with real (and nominal) interest rates as a result of the demographic trends you discussed?  Can Louis' argument that inflation is in structural decline due to technology hold up given this massive demographic shift that is coming?

    Reply
    • >Will.Denyer

      Will.Denyer

      20/10/17

      That is a fantastic piece.

      I use somewhat different age groupings in my analysis and GDP weight my measure and thus I get different conclusions in terms of timing, but I think that piece does a great job of exploring the potential implications of this demographic shift (higher real rates, higher real wages, and higher inflation). I agree with the first two; but I'm not sure on the last one. I plan to do more work on the implications for inflation, but here are a few ideas I'm throwing around in my head:

      1. If the Capital Providers Ratio starts to rise, and this means the natural rate also starts to rise, BUT market rates (whether policy rates on the short end of the curve or long rates) lag behind the rising natural rate, then the market rates will be below the natural rate and Knut Wicksell tess us this should cause inflation (either of capital goods or consumption goods or both). This is one potential interpretation of what happened in the 60s and 70s.

      2. It may be that another demogrpahic ratio is more useful in determining inflation. Perhaps it is the working age population ratio (say 20-64/rest) that matters more for consumer price inflation. The logic here would be that if the working age population is contracting relative to the rest, then the amount of goods and services that can be produced is falling relative to the demand for consumption. This could drive up consumer prices. This would at first seem to help explain what happend in 1965-1980, when US CPI and the global working age ratio rose, but it broke down thereafter--perhaps because Volcker hiked rates above the natural rate (going back to explanation 1). 

      3. Perhaps it is not demographic ratios that matter so much as population growth rates, or the growth rates of specific parts of the population (perhaps of those that demand a lot of investment and/or consumption?).  

      I plan to do more work on how demographis relate to inflation. As you can tell, I've been thinking about it, but I haven't come to any strong conclusions yet.

      I hope my initial thoughts are at least thought provoking. Sorry they are not better formed at this point.

      Will

      Reply
  • >Anonymous

    Anonymous

    18/10/17

    How does this factor in America's (and others') millenial generation, which is larger in size than the boomers?  Won't they be entering peak earnings/saving phase at the same time that the boomers' investing is accelerating, more than enough to offset?

    Reply
    • >Will.Denyer

      Will.Denyer

      20/10/17

      Dear Reader,

      Good question. However, according to my analysis they (I would say we, as I'm one of the older millenials, born in the early 80s, but I now live in Hong Kong) are not big enough to change the tide it seems.  The millenials are just starting to enter the net capital providing sweetspot (35-64), and yet the US capital providers ratio will fall in the coming years, and then more or less stabilize. See charts below.

      Best regards,

      Will

      Reply
  • >Anonymous

    Anonymous

    18/10/17

    This is simply a superb analysis! Thisis monumentallyu illuminating! Thank you, Will!

    Reply
    • >Will.Denyer

      Will.Denyer

      20/10/17

      Thank you for the very kind words. I'm glad you found it useful.

      Best regards

      Will

      Reply
  • >Anonymous

    Anonymous

    02/11/17

    This is the best I have read in a long time. I hope you will dig deeper and expand on the ideas you shared. I have not read the BIS paper mentioned yet, but inflation can be modeled with demographics and Engel's law, as many have done to explain the (goods) price booms of the 70's and 00's. But what about inflation in an economy where supply and demand are made up mainly by services and how can you invest in that if raw materials are not a key component any more?

    Reply
    • These are good comments. In general I think we need to do more work on the implications of the services-dominated economy, for productivity, inflation etc.

       

      Reply
    • >Will.Denyer

      Will.Denyer

      03/11/17

      I do indeed need to do more work on how demogrpahics will affect inflation, but my current thinking is this:

      If demographics start to push up the natural rate, then what this means for inflation depends on how policy and market rates respond. IF the policy and market rates lag behind then Wicksell would tell us that is inflationary (and that may indeed be what happened in the 60s and 70s). If however, policy and market rates keep up with inflation then this need not be inflationary. And if policy and market rates rise faster than the natural rate it could actually be disinflationary. 

      Does this Wicksellian framework still apply to today's more modern economy? My hunch is to say yes, generally it does. However, if we do have major technological breakthroughs this can effectively create savings out of thin air--by enhancing the value of the capital base without having to sacrifice consumption. If technogical advancements pickup then this could counter some or all of the demogrpahic headwinds.  

      Will

      Reply
  • >Anonymous

    Anonymous

    07/11/17

    Interesting paper Will but,  the "Savings Glut" view of economists such as Ben Bernanke and Paul Krugman needs to be vigorously rebuked. This incorrect view which was put forward to attempt to explain the Great Financial Crisis (GFC) by the main culprits and was challenged by economists at the Bank for International Settlements (BIS), particularly in one paper by Claudio Borio entitled "The financial cycle and macroeconomics: What have we learnt?":

    "The core objection to this view is that it arguably conflates “financing” with “saving” –two notions that coincide only in non-monetary economies. Financing is a gross cash-flow concept, and denotes access to purchasing power in the form of an accepted settlement medium (money), including through borrowing. Saving, as defined in the national accounts, is simply income (output) not consumed. Expenditures require financing, not saving. The expression “wall of saving” is, in fact, misleading: saving is more like a “hole” in aggregate expenditures – the hole that makes room for investment to take place. … In fact, the link between saving and credit is very loose. For instance, we saw earlier that during financial booms the credit-to-GDP gap tends to rise substantially. This means that the net change in the credit stock exceeds income by a considerable margin, and hence saving by an even larger one, as saving is only a small portion of that income." - source BIS paper, December 2012

    Their paper argues that it was unrestrained extensions of credit and the related creation of money that caused the problem which could have been avoided if interest rates had not been set too low for too long through a "wicksellian" approach dear to Charles and yourself.

    Borio claims that the problem was that bank regulators did nothing to control the credit booms in the financial sector, which they could have done. We know how that ended before. "Financing" doesn't equate "Savings".  If NIRP/ZIRP is killing "savings", there cannot be "proper" financing" to the "real economy" because the only fun place where the money has been flowing has been uphill to the bond markets, not so much downhill to the real economy. 

    Best;

    Martin 

    Reply
    • >Will.Denyer

      Will.Denyer

      10/11/17

      Dear Martin,

      Thanks for sharing. I reviewed this work by Bario. I completely agree that we have to consider developments in the "money economy" as well as the "real economy", manufactured "purchasing power" through money creation as well as real savings...., particularly when assessing cyclical matters.

      My assertion is that demographics appear to have exerted downward pressure on natural rates since the early 80s. In turn, I think market rates have also experienced a downward influence from this same demographic force. However, to Bario's point, we cannot assume that the natural rate and market rate fell in lock-step. The market and/or central banks surely drove market rates down faster or slower than natural rates at various points in time, with signficant consequences on the real economy. As you say, this is where the more cyclical Wicksellian Spread analysis comes into play.  

      I have no idea whether market rates today are above or below the natural rate.  I take Bario's point that low inflation readings are not necessarily proof that market rates are not unnaturally low--they might still be, and may be causing all kinds of distortions. 

      But again my view here is a more structural one. The downward influence of demographics on interest rates over the past 35 years now looks to be reversing.  As such, investors should pare duration on fixed income. 

      Best regards,

      Will

       

      Reply
  • >Anonymous

    Anonymous

    07/11/17

    Dear Will,

    Please have a look at this very interesting article:

    https://bankunderground.co.uk/2017/11/06/guest-post-global-real-interest...

    Best,

    Martin

    Reply
    • >Will.Denyer

      Will.Denyer

      10/11/17

      Martin,

      Very interesting article indeed. Their observation that periods of depressed interest rates tend to be followed by pretty hasty and aggressive rebounds in rate is sobering and has further encouraged me to reduce duration today.  

      Best regards,

      Will

      Reply
  • >Will.Denyer

    Will.Denyer

    10/11/17

    Dear All,

    In case you missed it, please note that I refined the CPR analysis, as described in our recent Monthly - How Much Longer For Low Rates?

    This refinement has resulted in a somewhat modified conclusion. As before, I come to the conclusion that the demographic tailwind for asset values has abated. But I now worry that demogrpahics are already starting to become a headwind (where as before it appeared we might have a 10-15 year period of relative calm before demographics turn for the worse).  Thus, the new analysis has me even more cautious on risk assets, and recommending an even shorter duration on fixed income. Allow me to clarify:

    In the simple CPR analysis, I relied on a simple binary ratio (similar to the dependency ratio, or working age ratio, etc…). Those between the ages of 35-64 were considered to have a net propensity to provide capital to the world capital market. Those  outside that age group were considered to be net demanders/borrowers of capital. Thus I had a simple ratio of 35-64 year olds/rest of population.

    But as the first chart below suggests the propensity to export capital is strongest the middle of that 35-64 year age cohort, and is almost neutral near the ages of 35 and 64. Similarly, the net demand for capital varies in strength throughout ones youth and retirement ages. The new Refined CPR adjusts each country's "excess saving" propensity according to much more granular changes in age structure. The result is the somewhat worrying 2nd chart below:

    Will

    Reply