Preparing for the worst

As Europe plunges ever-deeper into crisis and the U.S. limps between recovery and recession, investors are asking how much more can they take. They won’t know till they’ve imagined the worst. In other words, they won’t know until they’ve submitted to a stress test.

For banks, stress-testing means determining how much they are likely to suffer in adverse situations. For example, Canadian banks are modelling pressures on their regulatory capital should the housing market drop in value by 25% (it has happened before).

Stress tests gauge the capacity of their Tier-1 capital—the bank’s liquidation value for shareholders— to absorb losses. Banks are inherently leveraged institutions, with less than 10% of capital backing assets such as mortgages or derivatives.

“You have to be prepared for some extreme scenarios, even if you’re not ready to assign a probability to them,” notes Moshe Milevsky, a finance professor at York University’s Schulich School of Business.

“An economist [doesn’t know] the probability that the euro will break up or that Greece will go back to the drachma because those events have never happened before.”

Ideally, institutions with sophisticated portfolio management programs can create their own scenarios for stress-testing rather than simply relying on historical scenarios.

Testing for retail clients

Stress-testing is different for retail clients.Unlike institutional investors, they don’t actively trade their portfolios. In addition, extreme portfolio events can throw a retirement plan off track.

Unlike a pension plan with many contributors, all of whom have different retirement dates, there is only one contributor to a retail plan. So there’s limited scope for continuing contributions to make good the losses, short of a stepped-up contribution rate. Which makes stress-testing that much more important: to forestall depleting a retirement portfolio ahead of schedule.

Stress Testing part 1

Portfolio building begins with these fundamentals:

• Initial wealth
• Target wealth and target date
• Projected contributions on savings
• Projected return on savings

The most basic stress test concerns risk tolerance:

• What level of market losses is acceptable
• How that level affects the contribution rate
• How market losses will affect expected retirement date

For Milevsky, accounting for extreme events means “pensionizing” some retirement assets. “People should go out and get a pension if they don’t have one, particularly because for these worstcase scenarios,” he says. This means purchasing annuities and guaranteed lifetime withdrawal benefits in addition to stocks and bonds.

There are at least three elements in a retirement plan: a target income level, a savings rate and an expected return on savings. Stress-testing concerns the last element. Advisors can use many approaches to assess the risk to expected returns, and they are frequently complementary: stress is, after all, the ultimate edge of the risk spectrum.

To do a proper risk assessment, it’s not enough to rely on a verbal description of a client’s risk appetite. Instead, risk has to be quantified: how the client would feel in the event of a 50% drawdown in the market.

That’s the approach a Mississauga, Ont.-based advisor with a large brokerage uses. “Since my portfolios are already as broadly diversified as possible, rather than doing exhaustive stress-testing of my portfolios, I spend more time trying to get a good idea of my clients’ risk tolerances and educating them about investment risk. I’m stress-testing my client’s ability to take financial risks.”

Risk tolerance is important because of its behavioural impact. Drawdowns may cause investors to leave the market, missing out not only on the possibility of recovery, but also on the upside of the equity markets.

As Thornhill, Ont.-based financial planner Jim Otar points out, the extreme equity market returns and losses are generated on exceptional days.

Since 1900, such days have occurred only 6% of the time. So there is a very narrow window in which to reap the market return, and as Otar notes, most investors underperform the market precisely because they don’t stay the course.

“Investors try to time the market, they get scared, or they get carried away,” he says, citing Dalbar’s research on how investors buy high and sell low. “The actual performance of the individual’s portfolio is much worse than the average return in many cases.”

The opportunities

But risk is also opportunity. There are some advisors using institutional- level platforms, says Mike Archibald, senior institutional account manager at CPMS, a division of Morningstar Canada.

CPMS’s Backtest software examines Canadian and U.S. stocks across 4,000 data points, similar to what banks would do with their stress tests.

Advisors can also use stress-testing to find out the performance of the 50 cheapest price-to-book stocks at various points over the past 20 years, for example.

“We typically ask clients [about their] investment processes,” he says. “Let’s assume you’re a dividend-income-oriented investor. We would run [their chosen] portfolio on a monthly basis through the back-test.”

At the end of the process, clients are able to see the sensitivity of their stocks to various factors, such as valuation, growth and income. They can also see the effects of a +/- 20% fall in the market, for instance, compared to a benchmark like the S&P/TSX 60. But it may not be necessary to go that far.

“For many years, as I’ve counselled advisors on portfolio construction, the one thing I’ve suggested is to assume that a particular portfolio’s maximum drawdown will be roughly equal to the equity allocation times 50%,” says Dan Hallett, Director of Asset Management at Highview Financial Group in Oakville, Ont.

“The idea [is] that stocks get halved every once in a while (seven-to-nine years on average). And this is the temporary downside that the client must be comfortable with.And hopefully this is consistent with the advisor’s profiling work.”

What’s wrong with Monte Carlo

Stress Testing part 2

The projected return on savings can be affected by:

• Unexpected or “Black-Swan” events (e.g., the tech meltdown in 2000 or the financial crisis of 2007-2009)

• An unexpected sequence of bad returns that occur just before retiring

For Black-Swan events, you can model at the individual security or portfolio level historical or potential shocks to, among others:

• equities
• bonds
• interest rates
• inflation

Other factors can be added, particularly for individual securities, depending on the sophistication of the software used.

Milevsky notes many advisors have become comfortable with Monte Carlo simulations. Such a simulation might run through 1,000 random draws to determine the probability of a retirement portfolio’s success. An advisor might be able to determine that at a given level of savings and volatility in returns, a portfolio has an 80% chance of success.

But that can be deceptive. “When you do a Monte Carlo, you don’t know how sensitive your number is to your assumptions,” says Milevsky. “That’s where people have to do more work on the sensitivity analysis.When someone tells you there’s a 90% chance you’ll be okay, it sounds comforting, but that 90% chance rests on the assumption that interest rates are 4% a year. They’re not. Right now they’re 0%.”

That’s not to say that Monte Carlo simulations are wrongheaded, as some academics have proposed.

“Monte Carlo simulations based on normally distributed returns can be modified to account for fat tails, or excess kurtosis,” he says. “As long as you’re willing to assign a probability to an event and talk about its magnitude, you can always throw it into a Monte Carlo.”

More problematic is the sequence of returns. A bad return—such as 2008—depletes the capital that can be drawn down, potentially delaying retirement.

Thus, to assess the adequacy of retirement plans in the face of extreme events, Milevsky suggests basing the projection not on the 80% of simulations that are positive, but on the 1% that are the worst. An advisor would do this for a three-year period—what he calls a Black-Swan scenario.

“I think that’s what risk management on the personal level is all about. There are certain events that we just can’t quantify, but you want to make sure you’ll be prepared. I think sensitivity analysis on Monte Carlo is one way to get there.”

The sequence of returns is a theme Otar has also been working on for a long time, as much for behavioural as for statistical reasons. But he uses a 10%threshold for unlucky outcomes.

“Starting from engineering principles, you want to have a failsafe system,” he explains. “Planning for absolute non-failure is very onerous. You really have to go down to 3% withdrawal rates, and most people can’t. I allow for 10% probability of depletion at the beginning of retirement. If things go wrong, that gives the advisor enough time to switch to other methods of creating income, such as life annuities.”

Bell curve not adequate

Stress Testing part 3

For a sequence of bad returns, you can model historical returns or simulate potential returns using:

• Monte Carlo simulations
• actual historical sequences

Again, there are commercial programs that address this.

No test will be completely successful; the future is as yet unwritten. Nevertheless, advisors can get a good sense of the probability of portfolio success using historical-return data or simulations. This may mean dialling down the risk by adding bonds, boosting the savings rate, or examining when to pensionize or annuitize some of the retirement portfolio.

But Otar doesn’t use forecasts or Monte Carlo simulations. He instead uses historical outcomes, which he calls “aftcasting.”

One problem withMonte Carlo simulations is that they dampen the impact of the best and worst 3& of outcomes because they are based on a Gaussian analysis—the bell curve, or normal distribution. Normal markets—the Gaussian market—occur 94% of the time, he says.

“If it’s 94% of the time, why not use Gaussian?” asks Otar. “The damage is done in those 3% bad times. Or in those 3% good times, people get carried away and put too much money into the market.”

Under a normal distribution, Monte Carlo simulations assume events occur randomly. But, Otar points out, the historical sequence is not random: down years are followed by down years, and up years are followed by up years in long cycles.

“I take the actual year-by-year market history performance, inflation, bonds and interest rates and then calculate what would the portfolio do under those circumstances,” he says. “When you have simulations, you lose the sequence of returns.”

The “aftcast” of all years since 1900 shows something else: The role of equities in achieving portfolio success is overestimated. For a 30-year horizon, the optimal allocation is 40% equities and 60% bonds.

Indeed, as evidence, one may point to the fact that as of 2011, U.S. bonds have yielded a better return than U.S. stocks over the past 30 years—a result that doesn’t sit well with randomized simulations or modern portfolio theory.

“The events that happen so rarely are what affects the performance, not what happens most of the time,” Otar says. “If you want a maximized portfolio, you have to go to 40% or 50% equities. That’s to protect your money on the downside and have some money to put into the market when it’s down.”

However you perform a stress test—using stocks, a risk questionnaire, or simulations—one conclusion is clear: investors probably need more bonds in their portfolio than they think, even if they have long horizons and an aggressive risk tolerance.