13

Stocks Versus Bonds and Something About Precision Weapons

Utility theory is meant to represent someone’s tolerance for risk.

—JPP

WE’VE JUST DISCUSSED WHY WE SHOULD NOT GIVE UP ON asset classes. Though the debate on asset classes versus risk factors is relatively recent, the perennial, most fundamental debate in asset allocation is on stocks versus bonds. How much should an investor allocate to each asset class? This decision, more than any other portfolio construction decision, drives the risk level in investors’ portfolios. Ultimately, it drives outcomes with respect to an individual investor’s retirement security (or lack thereof) or ability to meet any kind of liability. It is, without a doubt, the most important portfolio construction decision an investor makes.

While I worked at a bond powerhouse (PIMCO), I drank the fixed income Kool-Aid. I often ended my presentations with the exclamation: “Bonds are awesome!” To assuage concerns about rising rates, I would emphasize how bonds diversify portfolio risk and hedge liabilities. Bond cash flows are known in advance, so you know how much you’ll be able to spend in the future. You can even hedge inflation risk with Treasury Inflation-Protected Securities (TIPS).

Now that I work at a stock-picking powerhouse (T. Rowe Price),¹ I often declare: “Stocks are awesome!” To assuage concerns with high valuations and short-term exposure to loss, I explain how a substantial allocation to stocks is the only way to reach long-term retirement goals when an investor can’t contribute enough. Stocks can provide an upside that can’t be achieved with bonds. They’re the engine of growth in investor portfolios.

Clearly, the question is not as simple as one versus the other. There’s a role for both stocks and bonds in balanced portfolios. Investors must calibrate their stock-bond allocation based on their risk tolerance. But how do we determine the appropriate mix, given an investor’s risk tolerance? Or an even more difficult question: How do we estimate an investor’s risk tolerance? The so-called glide paths used in target-date funds (TDFs) provide a useful guide to adjust the stock-bond mix as a function of when an investor expects to retire.

In a 2016 InvestmentNews editorial, I officially turned my coat on my PIMCO days. My colleague Jim Tzitzouris, an industry thought leader on life cycle investing, and I argued that many investors may need a higher allocation to stocks than they think. Jim had convinced me through a series of fun and intense debates over afternoon coffees (Jim drinks “quad” espressos, i.e., four shots). To make our case, we discussed personal finance and the role of human capital.

Some Background on Personal Finance²

At a 2006 Boston University conference, Nobel Prize–winning economist Paul A. Samuelson asked the audience whether personal finance was an exact science. Then he answered his own question: “Of course, the answer to that is a flat no. If this disappoints anyone in the audience, now is a good moment to rectify your miscalculation by leaving.”³

One of Samuelson’s many contributions to the field of economics has been to build mathematical models to better understand how to optimize personal finance decisions. How much should individuals save for retirement? How should they allocate their investment portfolio throughout their lifetime?

Samuelson’s models have evolved over time. Expanding on them, Bob Merton, Zvi Bodie, and others have argued that human capital—the present value of an individual’s future salary income—is an “asset” just like stocks and bonds and should be considered as part of the life cycle asset allocation decision.⁴

Human capital also drives the retirement liability. When they reach retirement, individuals need an income stream to meet their spending goals, usually some portion of their salary just before retirement. With defined benefit (DB) plans, this liability is explicitly defined; accordingly, a majority of DB plans employ liability-driven investment strategies. This approach has also been making its way into portfolio construction models for target-date funds, which are used in defined contribution plans.

Overall, the concept of human capital makes a lot of sense. However, Jim and I argued that the conclusion reached by most industry participants regarding its impact on portfolio construction is wrong.

Because salary payments are relatively steady month to month, conventional wisdom is that human capital is bond-like, and therefore, it should be hedged with bonds. In recent years, Bodie has even gone so far as to suggest that most individuals should invest 100% of their retirement savings in a TIPS portfolio, to safely match their retirement spending goal.

Is that sound advice? Echoing Mr. Samuelson: the answer is a flat no. While bonds will always have their place in balanced portfolios, ultimately Jim and I explained that individuals are more likely to reach their retirement goals with stocks. This conclusion leads to a key tenet of portfolio construction: in addition to risk tolerance, investors’ goals should guide their stock-bond mix.

The asset-side case for stocks is unambiguous. A recent study published in 2019 reported that only 10% of Americans are confident that they’ll have enough put away for retirement.⁵ People are severely underfunded, and the media refers to this situation as a retirement crisis.

With 30-year nominal bonds currently yielding 2.86% (and 30-year TIPS currently yielding 0.96%),⁶ bonds will not bridge the underfunding gap for individuals. In such a low-rate environment, the benefits of compounding are too small. Suppose TIPS generate 0% real return (lower than 0.96%, but bear with me for this example), and a couple save 10% of their consumption per year to buy TIPS. In this zero-rate environment, it will take the bond investor roughly 10 years to save enough to replace 1 year of consumption in retirement, assuming salary remains constant (or 9 years if we replace postsavings consumption). If you’re healthy and expect to live 30 years in retirement, you need to save for 270 years before you can retire (9 years of savings for 30 years of consumption postsavings). While this example assumes you earn nothing on bonds after inflation and therefore get zero benefit from compounding, until yields increase, reality is not that far off. With such low rates, investors can’t afford a 100% bond allocation. The math doesn’t add up.

This example illustrates how ultra-accommodative monetary policy has pushed investors toward risky assets. It explains the thirst for yield that has compressed corporate spreads and led to increased demand for high- dividend stocks. In general, it explains the demand for stocks and the structurally higher valuations of US and global stocks over the last two decades.

In contrast, in countries with higher rates and a lower equity risk premium, portfolio construction focuses on bonds. Recently, on a trip to Brazil, I met with several institutional investors who explained why they hold very little, if any, stocks: in the past, they’ve been able to meet their return target with short-term government bonds. From 1999 to 2019, the short-term interest rate in Brazil averaged close to 15%.⁷ Of course, these high rates have come with periods of hyperinflation, but real rates have been high as well, and ultimately, the major Brazilian investors and wealth managers I met with still hold less than 10% in stocks.

Human Capital: Does It Look Like a Bond or a Stock?

Most people think that bonds hedge human capital better than stocks. Salaries look like bond coupons because the amount and timing of paychecks are known in advance. As we approach retirement, our human capital depletes. If I ever retire, it will probably be 20 or 25 years from now. In theory, I have plenty of human capital left, as long as my health holds out. My total portfolio contains a large allocation to an “asset” that represents the present value of my future paychecks. But one year from retirement, I will have almost no human capital left. If my risk tolerance doesn’t change, I should replace my human capital, as it depletes, with similarly behaved bonds. This way, my total portfolio allocation, and ipso facto its risk profile, does not change.

However, this approach is too academic. First, it assumes constant risk aversion. Yet it seems to me that most people become more risk-averse as they approach retirement and as the balance of their savings (their “pot”) grows.

Next, Jim and I argued that human capital is more stock-like than bond-like. When stocks do well, salaries tend to increase due to earnings growth (and vice versa when stocks do badly). From that perspective, human capital has a “positive equity beta.” In the asset management and banking industries, a significant portion of our income is incentive compensation. And for many in these industries, job security seems linked to the economic cycle. Our “equity beta” is quite high. Employees and entrepreneurs in cyclical industries (housing, materials, transportation, etc.) seem to be in the same boat. At the other end of the spectrum, perhaps tenured professors, medical professionals, and government workers have lower “equity beta.”

Although we recognized that exposures should and do vary across industries, Jim and I guessed that aggregate wage data would reveal significant equity exposure. We calculated the correlation between wages and the returns on stocks and bonds. To avoid issues with short-term volatility in financial assets, we focused on rolling three-year return correlations. We found that, net of inflation, the average-wage index was more correlated to stocks (+53%) than to bonds (+30%), based on data from 1952 to 2014. Hence, if we hold risk aversion constant, stocks may be the better replacement for human capital as it depletes.

Of course, stocks have greater exposure to loss than bonds, and their volatility can keep retirees up at night. As individuals approach retirement, they can mitigate this downside risk with bonds. But in the final analysis, which is the best long-term retirement portfolio? Is it bonds or a balanced portfolio with a healthy allocation to stocks? In our InvestmentNews article, Jim and I echoed Samuelson. We said, “We believe the answer is a balanced portfolio with a healthy allocation to stocks. And if this conclusion disappoints any of our readers, now is a good time to rectify your miscalculation by moving on to another InvestmentNews article.”

Target-Date Funds as the Default Allocation

In the United States, as defined contribution plans have grown, we have given individuals responsibility for portfolio construction. Individuals are presented with a menu of investment options, and they must choose how much to allocate to stocks versus bonds. Then, within each asset class, they must choose how much to allocate between different strategies and sub–asset classes. These choices aren’t easy to make for those who are not investment professionals.

Recently, I discussed this topic with a financial advisor. He argued that it was unreasonable to ask individuals to solve their life cycle portfolio construction problem on their own. “Most people don’t have the expertise required to make these investment choices. Would your surgeon ask you to perform surgery on yourself?” he asked rhetorically. Most individuals simply don’t make a choice. Those involved with defined contribution plans in the United States know that inertia seems to be the most powerful force that drives portfolio construction. The do-nothing option is always the most popular.

Inertia has given rise to the importance of default options. Whether individuals elect to contribute a percentage of their salary toward their retirement or, as is often the case, they are automatically enrolled by their employers, they don’t construct their portfolios; instead their employers (who are their plans’ sponsors) must decide how to allocate the funds.

For many years, the default option was cash. But this default has led to poor returns. If an individual has a long time horizon before retirement, it’s hard to argue that cash is a good investment. In fact, it may be the worst choice. It’s not even the “risk-free” choice, because cash needs to be reinvested along the way, at uncertain rates. We can’t predict the cumulative return for cash over multiyear horizons. In theory, the “safest choice” is a long, inflation-protected bond that delivers cash flows that match what we want to spend in retirement.⁸ Such an asset does not exist, but long bonds are typically used as a proxy. However, long bonds, as we have just discussed, are also a suboptimal choice when rates are low, especially when individuals are underfunded (and they can’t afford risk-free).

Stocks are awesome. But they also expose investors to significant short-term losses. How much should individuals allocate to stocks throughout their life cycle? That is the key portfolio construction question. As mentioned, it affects investment outcomes in a way that no other portfolio construction decision does.

Following the Pension Protection Act of 2006, plan sponsors can automatically enroll employees for monthly contributions to their retirement plan. They can use a multi-asset fund as the default option. These measures make inertia work in favor of employees. They can opt out or change their asset allocation, but if they do nothing (which is often the case), they automatically get exposure to a diversified portfolio with a healthy allocation to stocks. A popular default option is a target-date fund. A TDF starts with a high allocation to stocks when the employees (or “plan participants” in defined contribution jargon) are early in their career, and the allocation gradually shifts from stocks to bonds as they approach retirement (the glide path). Participants are assigned a TDF vintage based on their age.

As a disclaimer, our Global Multi-Asset Division at T. Rowe Price oversees over $250 billion in TDF assets (as of May 2019). We are the largest provider of actively managed TDFs.⁹ Here’s how research director Jim Tzitzouris introduces our approach to glide path construction, i.e., how we determine how much stock an individual should hold as a function of the person’s age:

We believe that investors’ utility (satisfaction) is derived from two sources: consumption and wealth. Investors’ levels of risk aversion, impatience, and wealth affinity lead to different glide paths. We choose a glide path that maximizes utility for as many plan participants as possible.

Utility is essentially the satisfaction an individual derives from various levels of income and wealth in retirement, given the person’s tolerance for risk. It’s a measure that relies on probabilities. To measure expected utility at each point during an individual’s life cycle, we need to calculate the present value of multiple, probability-weighted future investment outcomes. Each investment outcome is mapped to a utility score. To generate the scenarios, we use Monte Carlo simulations, which embed forward-looking, long-term capital markets assumptions on the equity risk premium, interest rates, inflation, etc. (We will discuss direct utility maximization, also called “full-scale optimization,” in Chapter 14.)

This approach sounds technical, and it is, but we can summarize it as follows: We design the glide path in a way that considers what individuals are trying to achieve and how much risk they’re willing to bear in order to get there. It’s the good old return versus risk portfolio construction process, but across multiple time periods and with multiple objectives.

Our “utility lab” has been developed over more than 15 years, and our portfolio construction is difficult to replicate as a simple formula or recipe. Some of the objectives and constraints are “behavioral” in nature in that they don’t fit neatly within classical utility functions. Often, they represent plan sponsor needs. Also, our portfolio managers and researchers use judgment and experience in combination with data and models. They adjust portfolio weights on the margin to account for the uncertainty in the models and specific market needs. It’s an iterative process.¹⁰

Separately, we also optimize allocations to sub–asset classes and strategies. This second optimization process is how we “fill the buckets” within stocks and bonds. To do so, we use a variety of portfolio optimization techniques, scenario analyses, and, again, judgment and experience.

We’ll discuss portfolio optimization in more detail shortly. But for now, in the absence of any other information (such as taxes, other sources of income, goals, bequest motive, risk tolerance, etc.), what allocation of stocks versus bonds would we deem “best practice” as a default option for individuals? We have two different TDF series, one with higher stocks allocation than the other. Our higher-stocks glide path is by far the most popular. It has a longer track record, and it has delivered higher returns over time, because, well, stocks are awesome. However, there’s a role in the marketplace for a lower-stocks glide path, for plan sponsors and participants who are more averse to short-term losses and are better funded. These participants may have higher account balances, as well as higher savings rates, and perhaps access to a defined benefit pension.

Figure 13.1 charts our stocks versus bonds mix for our flagship, higher- stocks glide path, as a function of how many years the individual is from retirement (negative numbers show years in retirement). As shown, when an individual is 25 years away from retirement, allocation to stocks is 90%. The time horizon is quite long, so risk tolerance is higher. This allocation then gradually decreases as time passes. The at-retirement allocation to stocks is 55%, which may seem high, but again, most individuals are underfunded and need a nest egg to keep up with inflation and last for 20+ years in retirement.

FIGURE 13.1 Retirement glide path

To summarize, how much to allocate between stocks and bonds is the most important asset allocation decision. And it often requires a multiperiod optimization. We invest for individuals throughout their life cycles, and they spend over multiple periods in retirement. Defined benefit pension plans use an asset liability approach. They define the risk-free asset as the liability- matching portfolio. But if we think about the goal of defined contribution plans in the United States, and retirement-oriented plans throughout the world, the goal is always the same: to meet future income needs in retirement. To me, it’s a puzzle that different investors around the world use different approaches to solve the same problem. I pontificated about this issue at a pensions and investments conference back in 2016.¹¹ Endowments and sovereign wealth funds may seem like they have different objectives from each other, but essentially both entities invest to meet future cash disbursements.

The level of funding matters. It’s easy to simply recommend, “Match the liability,” or a favorite of consultants, “Match the liability with bonds, and then invest the rest in risk assets.” However, when expected returns on bonds aren’t high enough, you can only match a liability if you’re very close to fully funded. As an extreme example (just to make the point), suppose I have $1,400 in savings, and I want to go on a vacation that will cost $2,000, five years from now. And suppose real interest rates are 0.5%. How can I “match the liability” with bonds? Impossible. Five years from now, if I invest in bonds, I’ll have $1,435. The only two ways forward are to either contribute more or take some investment risk and hope for higher returns. If I take some investment risk, I need about a 7.4% return annualized to reach my $2,000 target. Underfunded investors need high returns. When consultants say, “Match the liability with bonds, and then invest the rest in risk assets,” this advice can only apply to overfunded investors.

Once we have solved the stocks versus bonds question, we have (roughly) calibrated the asset mix to the investor’s risk tolerance. Next, investors must populate the stocks and bonds “buckets” with multiple asset classes and strategies. For these granular portfolio construction choices, I submit that single- period optimizations work best. Single-period optimization models are more flexible and transparent than the multiperiod utility framework, and they are reviewed more often (every one to five years). Think of the multiperiod utility lab as the heavy artillery and the single-period optimizations as lighter, precision weapons. (I’m clearly not a military expert, but you get the idea.) Note that I refer to portfolio optimizations, plural. For robustness, it’s better to test various methodologies than to rely on a single model. As I’ve mentioned before, judgment and experience must always play a role in portfolio construction.

In Chapter 17, we will populate the stock-bond glide path with sub–asset classes. We will review in detail the sample asset mixes that are broadly representative portfolios that are used to manage a total of more than $250 billion in retirement assets. Because retirement platforms in the United States are typically focused on US assets and cost minimization (hence, alternatives are rarely included), we will look at a less constrained, more global multi-asset portfolio.

Before we review the end product, let’s discuss how we can use single- period optimization tools in the process, as well as cover two key topics in portfolio construction: the role of private assets and the active versus passive debate.

Notes

1.   We also have a very strong fixed income franchise, but it is smaller than our equity franchise.

2.   This section is from Page and Tzitzouris (2016).

3.   See Horan (2009).

4.   See, for example, Bodie, Merton, and Samuelson (1992), Merton (2003), and Chen, Ibbotson, Milevsky, and Zhu (2006).

5.   https://www.cnbc.com/2019/06/27/how-many-americans-have-nothing-saved-for-retirement.html.

6.   As of May 8, 2019. Sources: Federal Reserve Bank of St. Louis and cnbc.com.

7.   https://tradingeconomics.com/brazil/interest-rate.

8.   For an intuitive description of how to think about the risk-free rate as a function of your time horizon, see An Economist Walks into a Brothel, and Other Unexpected Places to Understand Risk by Allison Schrager (2019).

9.   Morningstar.

10.   Jim Tzitzouris recommends the book Strategic Asset Allocation by Campbell and Viceira (2002) as a good reference for several of the methodologies that we use.

11.   See video.pionline .com/media/Different+approaches%2C+same+problem/0_ebr4ogkk.