Sunday, May 26, 2019

Forecasting Future Investment Returns

To help you develop a reasonable annual spending budget, we provide you with Actuarial Budget Calculators (ABCs) that employ default assumptions for future investment returns, future inflation and your expected lifetime planning period.  These default assumptions are selected to be approximately consistent with assumptions used by insurance company actuaries in pricing current inflation-adjusted life annuities (net of expense loads and profit).  Thus, the Actuarial Budget Benchmark (ABB), which uses the default assumptions, provides you with a lifetime spending plan that could theoretically be fully funded through the purchase of relatively low-risk inflation-adjusted annuities at current market rates (the market value of your future spending liabilities).  The default assumptions currently are:

  • 4% per year investment return (discount rate)
  • 2% per year inflation
  • A lifetime planning period based on a 25% probability of survival for a non-smoker in excellent health from the Actuaries Longevity Illustrator
As we have discussed many times in this blog, we don’t expect that you will actually try to defease (or mitigate) all the risks associated with funding your future spending liabilities by going out and purchasing inflation-adjusted life annuities (or even fixed dollar life annuities).  We understand that many of you will invest some or most of your investable assets in more risky investments, such as equities.  Even though equities are more risky than investment in annuities, you (and we) generally invest in them because historically they have earned higher real rates of return than fixed income investments, and we expect to earn higher real rates of return on more risky investments like equities than on less risky investments like annuities and other fixed income investments. 
 

So, if we invest some or all of our accumulated savings in equities and we expect to earn higher returns on them, why do we use these relatively conservative assumptions as the default assumptions in our ABC workbooks?  In this post, we will once again address this common question and point our readers to several sources that they may find of interest if they want to override our default assumptions when developing their spending budgets. 
 

Using our ABCs to develop a current spending budget data point
 

As discussed above, our ABCs utilize default assumptions to calculate a current year spending budget that is generally expected to remain constant in real dollars for the duration of retirement if all assumptions about the future are realized.  Since we don’t know what the future will be any more than you do, and since you may desire a pattern of future spending that is not necessarily constant in real dollars from year to year, we also allow you to override these default assumptions. 
 

The beauty of the Actuarial Approach is that it automatically adjusts for deviations between actual and assumed experience (including deviations from assumed investment performance) on an annual basis as part of the annual actuarial valuation process.  Therefore, if your risky equity investments earn more than 4% in the current year and you don’t smooth your results, your ABB for the next year will automatically reflect the current year actuarial gain on investments and will increase from the real-dollar ABB expected for that year, all things being equal.  Thus, if you are using the ABB to determine your annual spending budget, you are not losing actuarial gains from superior equity returns; you are simply deferring recognition of those gains until future years. 
 

Note that if you use the ABB to develop your spending budget and you invest a portion of your assets in risky investments, there is no guarantee that your spending budget will not decrease from one year to the next.  Each year’s return can be higher or lower than the 2% real rate of return assumed in the ABB calculations.  Therefore, while we (and you) may believe that assuming a 2% real rate of return is relatively conservative, and you expect to earn higher real rates on your investments in equities, you should know that your real dollar ABB can decrease from one year to the next depending on actual realized investment returns (and actual inflation and longevity).  The bottom line here is that if you invest in risky assets, you haven’t mitigated your investment risk, and your spending budget may decrease (or increase) from one year to the next if you don’t smooth it.

In addition to being consistent with basic financial economics, one of the primary reasons that we recommend using these relatively conservative annuity-based default assumptions is because we believe most of our readers would rather experience actuarial gains that increase their future spending budgets than actuarial losses that decrease future spending budgets.  If our annuity-based default assumptions are used, then gains from investment in equities will be recognized as they are earned and not “pre-recognized”, as is common with many Monte Carlo (MC) models that assume expected rates of investment return based on historical return data.

If you or your financial advisor are using a Monte Carlo model that indicates you have a high probability of being able to spend amounts significantly higher than the ABB, then you are probably “pre-recognizing” expected future investment gains (or you are not doing an “apples to apples” comparison by not recognizing the same assets or other spending liabilities).  If this is the case, we believe that the real rate of return assumptions being made for equity investments in the Monte Carlo model may be overly optimistic and may possibly not reflect the current market P/E environment (as discussed below).
 

Forecasting future investment returns
 

So, if you believe a 2% real rate of return assumption is too conservative, how does one go about developing a better future investment return assumption?  Research has shown that the current level of equity market P/E ratios appears to be a more important factor in predicting future equity returns than historical return data.  In the recently released practice note for pension actuaries entitled, “Forecasting Investment Returns and Expected Return Assumptions for Pension Actuaries” the authors conclude:

“Historical returns (and other capital market information) provide helpful data for thinking about future returns, but average returns from historical periods are not, by themselves, strong indicators of future returns. However, an analysis of historical data for components of return forecasts—yields, price/earnings (P/E) ratios, earnings growth relative to GDP growth, inflation, etc.—are used in conjunction with historical returns to understand how current levels of those factors might affect the individual components that make up an investment return assumption.”
 

and,
 

“Because P/E ratios are just the inverse of earnings yield (earnings yield = E/P) they provide a very useful indicator for future returns. Generally, the higher the starting P/E ratio (lower yield), the lower future returns are likely to be.” “The most common type of P/E ratio used for return forecasting is the Cyclically Adjusted Price- Earnings (CAPE) ratio introduced by Robert Shiller in 2000.”

The Case-Shiller PE Ratio can be found here and at the time of writing was 29.10.

In a November, 2012 article entitled, “An Old Friend: The Stock Market’s Shiller P/E”, Clifford S. Asness examined the correlation between P/E ratios for equities and subsequent returns over a ten-year period.  He concluded, “Ten-year forward average returns fall nearly monotonically as starting Shiller P/E’s increase. Also, as starting Shiller P/E’s go up, worst cases get worse and best cases get weaker.”  Dr. Asness’ paper presents a summary of 10-year S&P returns by starting P/E ratio for each year from 1926 through 2012.  For example, he indicated the following 10-year returns for years starting with P/Es between 25.1 and 46.1 (i.e., like the current P/E environment):


Average real 10-year return:     0.5%
Worst real 10-year return:       -6.1%
Best real 10-year return:          6.3%
 

By comparison, similar 10-year returns for years starting with P/Es between 13.8 and 15.7 were:
 

Average real 10-year return:    8.0%
Worst real 10-year return:      -0.9%
Best real 10-year return:        15.1%
 

While Dr. Assness’ data indicates that it is possible for 10-year real stock market returns in today’s 29 P/E ratio environment to exceed the approximate 2% return expected for inflation-adjusted annuities, the expected average real 10-year return (based on historical data) is less than 1% and the expected worst case real 10-year return is -6%.  Of course, these expected returns can be affected by many other factors, including changes in future interest rates (with increasing interest rates generally lowering returns), but on average, Dr. Assness’ data shows expected real equity returns for the current P/E environment to be less than historical average equity returns.  
 
The key takeaway from Dr. Assness’ paper for us (other than the fact that the current P/E ratio may disappoint those of us who expect higher future returns on risky equity investments), is that there are expected equity returns based on historical data averages and there are expected equity returns based on historical data adjusted for starting P/E ratios (and perhaps other factors).  If you or your financial advisor are using a Monte Carlo model to help you develop your spending budget (or for other reasons), make sure that expected equity returns assumed in the model are consistent with today’s market P/E ratio and are not based solely on historical averages. 
 

Conclusion
 

We continue to believe that our Actuarial Budget Benchmark (ABB) is a useful data point for helping individuals and couples develop a spending budget.  Even if you use the ABB, and its relatively conservative assumptions, however, there still is no guarantee that your spending budget may not go down in future years if you invest in risky assets and those assets fail to earn expected returns.
 

While Monte Carlo models may be very useful in assessing risks and selecting investment strategies, we remain unconvinced that these models are superior to the Actuarial Approach for purposes of developing a robust spending budget (contrary to the belief of most financial advisors).  We caution individuals to be very skeptical of claims that they may have a 92.7% probability of being able to spend $X per year if they invest Y% of their assets in equities. The reasonableness of MC results is very much dependent on the assumptions made for future asset returns.  We encourage you to look under the hood (and kick the tires) of a Monte Carlo model you use to make spending decisions to make sure the model reflects current equity market P/E ratios rather than simply uses historical average returns.  We also encourage you to compare the model results with your ABB to make sure your spending budget does not “pre-recognize” higher expected returns before they are earned.