The Next Generation of Income Guarantee Riders: Part 3 (The Income Phase)
December 11, 2012
by Wade Pfau
This is part three of a threepart series of articles reviewing standalone income (SALB) guarantees. Part one of this series is available here and part two is here.
What is the purpose of an income guarantee? Some view it as a traditional insurance product, meant to provide downside risk protection, while others like that it encourages clients to accept a more aggressive asset allocation, promising more upside without exacerbating worstcase outcomes. Can it really be both?
Some would argue that guarantees aren’t worth bothering with at all, since many clients can expect to do just as well without paying the extra fees.
So who’s right?
In this third and final installment in my series on guarantee riders, I’ll focus on the postretirement income supported by income guarantee riders for variable annuities (VA/GLWBs), standalone living benefit riders (SALBs), and an unguaranteed portfolio of mutual funds. I’ll highlight how differences among these products affect their end results, while also investigating what roles guarantees can most appropriately play in a retirement portfolio.
Setting the stage
For many advisors, it’s a welcome relief that clients no longer have to permanently commit their assets to a deferred variable annuity in order to purchase a rider that guarantees lifetime withdrawal benefits. The RetireOne product from ARIA is one such rider – it can be applied to a portfolio of mutual funds – and for this analysis, it is the one that I’ve focused on. The relatively lowcost GLWB rider available for Vanguard’s variable annuities, meanwhile, is what I’m using as a relatively attractive example of an annuity rider.
For a full refresher on the background and features for VA/GLWBs and SALBs, I’d recommend reviewing part 1 in this series. Very briefly, these products are designed to provide owners with downside protections, upside potential, and the opportunity to have remaining assets returned prior to death or as a death benefit. To accomplish this, these riders guarantee an income for life at a fixed withdrawal percentage of the initial assets. As long as the investor does not exceed the allowed withdrawal amounts, guaranteed withdrawals never decrease (in nominal terms – they may well decline according to inflationadjusted metrics), even if the account balance falls to zero. If the value of the underlying account increases enough (after accounting for any withdrawals and fees), a stepup feature kicks in to provide permanently higher withdrawal amounts.
I’ve simulated the incomephase performance of the Vanguard VA/GLWB, the RetireOne SALB, and an unguaranteed mutual fund portfolio using Monte Carlo simulations. Parts 1 and 2 of the series, linked above, focused on the initial deferral period and the crucial moment when the income guarantee kicks in, respectively, relying on historical data. This final article will shift gears to consider the results of my simulation and their implications for the income phase of retirement.
Potential biases in research methodologies
Income guarantees are complicated financial products, and it’s important to understand that the research published about them often makes assumptions that can present the guarantees in an overly positive or negative light. Let’s review some of the most crucial.
Underlying fees: A particularly common approach in such research is to compare the results of using income guarantees to drawdowns from unguaranteed mutual funds. Readers must pay careful attention to the fees that are assumed to underlie those guaranteed and unguaranteed funds, which should be consistent across the different products. (That’s why I compare the lowcost RetireOne guarantees with Vanguard’s lowcost VA/GLWB and with unguaranteed lowcost index funds.)
Asset Allocation: Clients with income guarantees will naturally feel more comfortable accepting an aggressive asset allocation, and ideally one should compare approaches using the asset allocations a client would actually choose with a guarantee and without one, rather than assuming a onesizefitsall approach.
Underlying Returns: For obvious reasons, assumptions about expected returns can dramatically affect the results. With better performance and lower inflation, guaranteed approaches will reap the benefits of their higher upside when compared with a lessaggressive, unguaranteed approach, but the unguaranteed approach will also be more likely to support the corresponding guaranteed withdrawals. In other words, there will be less downside risk for the guarantee to protect. With more pessimistic return assumptions, there will be fewer step ups and less upside, but the guarantee becomes much more likely to matter.
Time Period: Assuming a longer retirement horizon skews results toward guarantees, since over time it becomes more and more likely that the unguaranteed portfolio will run dry. Some researchers believe that the appropriate time period to investigate is remaining life expectancy, while others opt for something like a 3040 year retirement, aiming for the high end of realistic scenarios.
Spending Rates: Finally, one way to make a guarantee look better is to guarantee only part of a portfolio while assuming that total spending will exceed the guaranteed payout rate. Doing so causes both an unguaranteed portfolio and a partially guaranteed portfolio to deplete more quickly, but the partially guaranteed portfolio will still look better, since it continues to at least provide a minimal amount of income. In either case, however, income may fall well below the basic needs of a client, who should have been advised to choose lower spending rate from the outset. (To avoid this problem, I simply compare the withdrawal amounts supported by a guaranteed portfolio to an unguaranteed portfolio that attempts to replicate the same guaranteed payouts for as long as possible.)
In addition to minding the underlying assumptions, we must also be clear about what sorts of outcome measures are most appropriate. Clients who view a guarantee rider as an insurance product may consider the guarantee to be primarily a form of downside risk protection, in which case any analysis should focus on the worstcase outcomes. (This is the approach I took in parts 1 and 2.) But another justification for income guarantees that they encourage a client to choose a moreaggressive asset allocation with higher upside potential, in which case the focus may shift to demonstrating which approach enjoys superior average outcomes.
Data and modeling approach
Advisors who are familiar and comfortable with the 4% safe withdrawal rate ruleofthumb may see little need for an income guarantee. But market conditions today suggest that pessimism may be in order – what worked for yesterday’s retirees may not work for today’s or tomorrow’s. Interest rates are at historical lows and stocks are overvalued, at least according to historically reliable metrics like Robert Shiller’s CAPE.
While parts 1 and 2 both analyzed outcomes for rolling periods from the historical data, as we turn to considering retirement income, the primary basis for this article will be the results derived from Monte Carlo simulation. Table 1, below, provides the asset market assumptions on which those simulations were based. For the most part, I used current market conditions to guide the simulations, but near the end of this article I’ll discuss how the results change under more optimistic assumptions. (For more detail on how I obtained the figures you see here, see Appendix 1 at the end of this article.)
Table 1 

Asset Market Assumptions Based on Current Market Conditions 


Arithmetic 
Geometric Means 
Standard Deviations 
Correlation Coefficients 


Stocks 
Bonds 
Inflation 

Stocks 
4.8% 
2.8% 
20.0% 
1 
0.1 
0.2 
Bonds 
0.0% 
0.2% 
7.0% 
0.1 
1 
0.6 
Inflation 
2.5% 
2.4% 
4.2% 
0.2 
0.6 
1 







Summary Statistics for U.S. Real Returns and Inflation Data, 1926  2011 


Arithmetic 
Geometric Means 
Standard Deviations 
Correlation Coefficients 


Stocks 
Bonds 
Inflation 

Stocks 
8.6% 
6.5% 
20.3% 
1 
0.1 
0.2 
Bonds 
2.6% 
2.3% 
6.8% 
0.1 
1 
0.6 
Inflation 
3.1% 
3.0% 
4.2% 
0.2 
0.6 
1 
My simulations assume a 65year old couple who buys the guarantee at 65 and immediately begins to take income. For simplicity’s sake, their retirement date wealth is assumed to be $100, though the results are, of course, scalable. Since the analysis is assumes current market conditions, the payout rates at retirement for the 65year old couple, until both spouses are deceased, are 4.5% for the VA/GLWB and 3.5% for RetireOne.
For the VA/GLWB, the payout depends only on age, while the payout rate for RetireOne as depends on the current yield on 10year Treasury bonds. The payout rate is 3.5% if the Treasury yield is less than 4.5%, and it can increase to up to 5.5% if the Treasury yield exceeds 7%.
For RetireOne, after the guaranteed income begins, the benefit base is no longer determines the withdrawal amount. Instead, stepups in withdrawals occur whenever the revised payout rate (a calculation that involves multiplying prevailing Treasury yields by the remaining account balance) exceeds the previous guaranteed withdrawal.
Since I do not attempt to simulate future interest rates, I’ve assumed that RetireOne’s payout stays at 3.5% throughout retirement. While that assumption could bias results somewhat against RetireOne, but any such effect is likely quite small, since interest rates are currently much less than 4.5% and, as we’ll see, it becomes increasingly unlikely for the portfolio to reach new highwater marks as retirement progresses.
When it comes to determining the value of a guarantee, it’s important to always consider whether an unguaranteed portfolio of mutual funds would be able to replicate the guaranteed payments without experiencing wealth depletion. As I explained above, the asset allocation for an unguaranteed portfolio in any such comparison should be less aggressive, though the exact allocations will depend on a client’s preferences.