How Advisors' Biases Can Cloud Social Security Claiming Advice

A recent article by Dr. Derek Tharp on Kitces.com,“Why Delaying Social Security Benefits Isn’t Always The Best Decision,” and versions of it in other media outlets, including an op-ed in The Wall Street Journal, made a substantial effort to push back on the idea that delay is favorable.

The acceptance of the methodology Tharp proposes by many financial advisors in supporting comments, in reviews and other media is deeply troubling because it risks justifying actions that produce substantially negative outcomes for clients, which ultimately reflects poorly on the advisory profession as a whole.

I have two major issues with the article and the underlying process. First, it exhibits a self-serving rationalization bias. Second, it is internally inconsistent because it is a victim of a narrow-framing bias.

Rationalization Bias

Imagine a used-car sale. The car has a serious problem — metal shavings in the crankcase, indicating the potential for massive engine failure. At the same time, the buyer isn’t car-savvy and doesn’t know to inspect the crankcase oil before buying the car.

The buyer negotiates the seller down by a couple thousand dollars. The seller knows that discount isn’t nearly enough to cover an engine replacement, but doesn’t necessarily know if the shavings are bad enough that an engine replacement is certain. The buyer is short-term thrilled.

The expected utility wasn’t properly evaluated by the buyer because the buyer didn’t have the information or expertise to make a fair evaluation. The asymmetry of information and expertise benefited the seller at the buyer’s expense. If the seller disclosed the metal shavings and the potential impact on the engine, would the buyer make the same decision?

We have a similar situation with the Social Security claiming decision.The soon-to-be retiree who is making the decision to “buy” more Social Security benefits through delay seeks information from a variety of sources, and their financial advisor is often a key influencer in that decision. The Social Security decision is complicated, and its interaction with the rest of the retirement income plan is even more so.

At the same time, there is a parallel between the incentives of the car salesman and the advisor who glosses over the Social Security decision. Advisors most often are not compensated for the Social Security decision. In fact, advisors are disincentivized in two ways.

First, the decision creates friction. Most clients don’t want to delay, as is evidenced by actual claiming behavior. An advisor who actively works to change that mentality risks losing the client, or at minimum, failing to make a new client, which is a direct loss of the marketing dollar or time spent to bring the potential client to the door.

Second, delay costs portfolio assets, on which most advisors are compensated. If the path is easier, and the result is a near-term better outcome for the advisor, and the perception of a better outcome for the client, who wins?

Tharp uses expected utility theory at the level of the Social Security decision rather than at the overall cashflow level. That usage “rationalizes” behavior that creates suboptimal outcomes.

If expected utility theory were applied at the plan level instead of the individual decision level, then the outcomes of the analysis, especially after incorporating the tax benefits that delay creates for many clients, would be aligned with what the majority of the academic community has advocated for years — most people who are financially capable of delay should delay at least one benefit in the household.

I don’t believe Dr. Tharp is intentionally selling a lemon, but instead is trying to understand why people act the way they do and how early claiming could potentially be rational.

Narrow Framing

Tharp introduces a framework where he identifies a host of risks and arbitrarily assigns an additional discount rate to these risks. His list of factors and discounts for the Social Security decision of a hypothetical couple are as follows:

• Mortality Risk: +2.0% (John and Sarah do not have longevity in their family and are concerned about this)
• Sequence of Returns Risk: +2.0% (their desired $60k annual spending amounts to a 12% withdrawal rate, making them highly sensitive to early downturns)
• Regret Risk: +0.5% (John and Sarah expressed they would feel regret around policy cuts or dying prematurely)
• Health Span Risk: +0.5% (they are inclined to defer spending or even retirement without Social Security income)
• Spending Optionality/Flexibility Decline Risk: +0.5% (delaying Social Security could deplete their retirement nest egg, leaving little flexibility)
• Underspending Risk: +0.5% (John and Sarah's saving background suggests they may spend less without Social Security income to rely on)
• Longevity risk: –1.0% (John and Sarah would benefit from longevity risk protection if they delayed claiming)

He says: “Adding these together results in a client-specific discount rate of 8.5% — the 4% real opportunity cost from their portfolio plus another 4.5% after addressing risks that increase or decrease their discount rate.”

Internal inconsistency is created by a narrow framing bias. To expose that bias, we can attempt to apply the same framework to other retirement income decisions. Here is my attempt to do so, specifically for the decision on how to invest the client’s portfolio.

Investment decisions in general traditionally begin with the risk-free rate, then discount the expected future cashflows of the investment by that rate, plus any additional discounts specific to that investments expected cashflows to determine whether the risk premium is worth the risk and to what extent.

Let’s follow Tharp’s framework and try to apply the same logic he uses to evaluate the Social Security decision to the decision of whether to invest in anything above the risk-free asset. Here is his list of factors, which I have adjusted for the risks of a 60/40 portfolio, or left the same, but changed the reasoning for the discount:

• Mortality Risk: +2.0% (unchanged) (John and Sarah do not have longevity in their family, so their needs are most likely going to be met without taking additional investment risk)
• Sequence of Returns Risk: +3.0% (*1.5) (The potential drawdown for a 60/40 portfolio is at least 1.5x the potential benefit cut)
• Regret Risk: +0.5% (unchanged)  (John and Sarah expressed they would feel regret around making a decision to take risk and then get beat up by a bad market.) How much regret? Who knows. This assumption is as arbitrary as Tharp’s was.
• Health Span Risk: +1% (*2) (If people spend income, not portfolios, then investing in any growth asset over an income-producing asset or ladder would create further underspending)
• Spending Optionality/Flexibility Decline Risk: +0.5% (unchanged) (taking any risk presents a spending challenge because we never really know when markets will tank)
• Underspending Risk: +0.5% (unchanged)  (John and Sarah's saving background suggests they may spend less if they are focused on growing assets for the future)
• Longevity risk: –1.0% (unchanged) (John and Sarah would benefit from the potentially higher returns they may achieve if they live a long time)

I would suggest you try to arrive at your own numbers and rationales for the discounts for this exercise. Doing so will give you an appreciation for just how arbitrary the modifications are.

Assuming you accept my thought process outlined above, the client would use that discounting to decide whether or not to invest in the 60/40 portfolio. They would determine that the discount rate of 6.5% is considerably higher than the expected portfolio return above inflation (4.89% as outlined in the Kitces.com article) and we should recommend a portfolio solely consisting of short-term Treasurys.

I expect most advisors won’t agree with that conclusion. If you don’t, then you shouldn’t use the portfolio return as the discount rate for Social Security benefits, and you most certainly shouldn’t add additional arbitrary discounts on top.

A robust financial planning process should be consistent across decisions, whether the advisor is compensated for that particular decision or not. Good Social Security planning processes, and software by extension, incorporate all significant rules including spousal and survivor benefits, and allow inputs for critical assumptions such as the age at death. They stress test for varying life expectancy and for the possibility of benefit cuts to show the impact of the decision under various other potential futures.

Finally, they allow input of the discount rate. Depending on your perceptions of the safety of Social Security benefits, that could be government bonds, or if you are more in the camp that Social Security benefits carry some “default risk,” it could be a high-quality corporate bond portfolio. But it most certainly should not be the portfolio’s return, because the portfolio composition and resulting expected risk and return is a decision that should be built up through the same decision framework, rather than assumed to be pre-existing.

When these assumptions are properly made, we can get an idea of the “value” of various Social Security decisions, much like a discounted cashflow analysis gives us a reasonable approximation of the value of a security. This decision ultimately gets brought into the plan, which then allows us to build the portfolio around the Social Security decision. Ultimately, “bridge strategies” and more aggressive long-term investment strategies are the frequent result of a consistent framework.

If we as an advisor community are concerned more about our clients and the future of the profession, we owe the effort to assemble a process that consistently identifies and evaluates the metal shavings. Sometimes the car is worth buying anyway.

In the Social Security space, when using appropriate assumptions, we frequently see that early claiming makes sense for single men, people in poor health, the lower earner in a couple, people who can’t afford to delay and certain age-gap situations.

Discounting future Social Security benefits by a portfolio return, or by an arbitrary, custom-built discount rate, or justifying with expected utility theory applied only to the Social Security decision and not to other assets, leads us as advisors down a self-serving path that sacrifices client outcomes for the advisor’s benefit. Ultimately, we should be wary of those temptations and focus on consistent processes that drive better decisions — or at least allow a client to make truly informed decisions.

Joe Elsasser, CFP, is the founder and President of Covisum, a financial planning software company.