GK Guardrails Explained: How the 2006 Decision Rules Actually Work

Two retirees with identical portfolios can spend very differently and both be making the right call. One holds spending steady through every market storm. The other adjusts as the portfolio breathes.

The constant-spending picture, the one behind the 4% rule, is a useful planning convention. It is not a description of how anyone actually behaves. Real retirees notice when their portfolio drops and react. They notice when it grows and reward themselves. The interesting question is not whether to adjust but how.

This post walks through guardrails: the family of withdrawal rules that bake the adjustment into the strategy itself. Bellavia now offers two flavours of guardrails: a simple parametric version with explicit user-set bounds, and the canonical Guyton-Klinger 2006 version, where the thresholds are fixed by the original paper and the strategy makes its own decisions. Both are dynamic. They behave differently.

1. The need guardrails answer

The 4% rule frames retirement as a binary outcome: at the end of 30 years, did you have a dollar left or did you not? Anything else, e.g. running out at year 28 or finishing with $5 million or having to skip a vacation in year 12, gets compressed into a single success/failure flag.

That framing is convenient for research. It is also wildly out of step with how retirees actually live. In reality, a retiree watching the portfolio fall 30% in year 4 is not going to keep withdrawing the same inflation-adjusted dollar amount and pretend nothing happened. They will adjust. The question is whether they adjust systematically or by feel.

A guardrail strategy is just a written rule for how to adjust.

Guardrails replace the binary "did the portfolio survive?" question with a continuous "what level of spending is sustainable, given how the portfolio is performing right now?" That continuous answer is updated every year, by the strategy, on the retiree's behalf.

There are two natural ways to build the rule:

• Parametric: pick a target spending amount and two rate bounds. If the spending rate climbs above the upper bound, cut. If it drops below the lower bound, raise. You set the bounds. This is what most calculators mean by "guardrails."

• Algorithmic: take a paper that worked out which adjustments are most effective and follow it precisely. Don't pick the bounds yourself. The most cited example is Guyton & Klinger (2006). This gave us the term "decision rules" in retirement planning.

Bellavia now offers both. The rest of this post explains how each one works, and where they disagree.

2. Bellavia's parametric guardrails

The parametric guardrails strategy in Bellavia takes four inputs:

• Target annual withdrawal — the dollar amount you'd take in a quiet year
• Lower guardrail rate — if your withdrawal rate falls below this, you raise spending
• Upper guardrail rate — if your withdrawal rate climbs above this, you cut
• Real or nominal basis — whether the target dollar amount is held in inflation-adjusted terms or kept fixed in nominal dollars

The default settings is $40,000 target, 3% lower bound, 5% upper bound. It corresponds to a 4% initial withdrawal rate on a $1,000,000 portfolio with ±25% bands. Everything is configurable.

The mechanics

Every year, the strategy computes one number: what fraction of the current portfolio your target withdrawal would represent. Three cases:

• Current rate above the upper bound — the portfolio has fallen enough that the target dollar amount is now an unsustainable percentage. Cut: withdraw the upper bound × current portfolio instead of the target.
• Current rate below the lower bound — the portfolio has grown enough that the target is now an artificially small slice. Raise: withdraw the lower bound × current portfolio.
• Current rate inside the bounds — take the target dollar amount as planned.

What this looks like in practice

Consider a retiree with $1,000,000 on January 1, 1966 — about as bad a starting point as US history offers. The next decade is the stagflation era: high inflation, weak real returns, multiple bear markets. A constant-$40K withdrawer gets steadily ground down. A parametric guardrails withdrawer with 3%/5% bands adapts as seen in the following:

Figure 1: Parametric guardrails (3%/5%) vs the constant $40K plan, 1966–1995. As inflation erodes the real value of the constant withdrawal and the portfolio struggles to recover, the parametric strategy snaps the withdrawal down to whatever 5% of the (shrinking) portfolio is. By year 15 the parametric retiree is taking less than $30K in real terms; by year 25 they are below $25K.

A few features worth noticing:

• Cuts come quickly when the portfolio falls. By the early 1970s the parametric strategy has already snapped the withdrawal down because the rate has crossed the 5% upper bound.
• Raises come gradually in this cohort because the portfolio never recovers far enough to push the rate below the 3% lower bound — it takes until the late 1980s for the strategy to start lifting again.
• The strategy never runs out. Because each cut is taken as a percentage of the current portfolio, the portfolio tends towards zero rather than hitting it. The constant-$40K withdrawer would have failed in this cohort without the adaptation logic.

The cost is variability. Worst-case real annual spending in this cohort drops to roughly $24K which is a 40% cut from the planned $40K. That possibility is something the retiree needs to know about before they sign up for the strategy.

3. The canonical Guyton-Klinger rules

In 2006, Jonathan Guyton and William Klinger published a paper in the Journal of Financial Planning titled "Decision Rules and Maximum Initial Withdrawal Rates." It was a turning point in retirement-planning research because it was the first widely-read paper to argue that initial withdrawal rates can be substantially higher and they suggested 5.3% rather than 4%, if the strategy includes systematic spending adjustments.

The paper proposed four rules, which together are now usually called "the Guyton-Klinger guardrails" or "GK." Bellavia offers this canonical version of all four under the strategy name "GK Guardrails (Original 2006)."

The four rules

1. Modified Withdrawal Rule (WR). In a normal year, the strategy raises the withdrawal by the inflation rate. But if the prior year's portfolio return was negative AND the candidate inflated withdrawal would push the current rate above the initial withdrawal rate, the strategy skips the inflation bump. There is no make-up: a missed bump is permanently lost.

The WR is the rule that keeps spending from running away during prolonged sideways markets. It activates every year in the background.

2. Capital Preservation Rule (CPR). When the current withdrawal rate exceeds 120% of the initial rate, cut the dollar withdrawal by 10%. The CPR is the lower guardrail, the "things have gotten bad, time to spend less" rule.

The CPR is suspended in the final 15 years of the planning horizon. The reasoning in the paper: if you only have 10 years left, taking a 10% cut is not going to materially extend the portfolio, and the cost in lifestyle is no longer worth it.

3. Prosperity Rule (PR). When the current withdrawal rate falls below 80% of the initial rate, raise the dollar withdrawal by 10%. The PR is the upper guardrail, the "things have gone better than planned, take more" rule. There is no sunset on the PR: it applies for the full retirement horizon.

4. Portfolio Management Rule (PMR). Where do withdrawals come from inside the portfolio? GK's answer: a five-step priority order. Take from overweight equity first (sell winners), then from overweight fixed income, then from cash, then from remaining fixed income, then from remaining equity in order of prior-year performance. There is also an "equity skip" sub-rule: in years when any equity sleeve had a negative return, skip equity entirely if cash and fixed income can cover the withdrawal.

The PMR also handles rebalancing: each year, positive overweights get skimmed back to the target weights, with the excess moved to cash.

What's fixed and what's not

Everything in the four rules above is fixed by the paper:

Notice what is not in the user's hands: the band width (always 20% on each side), the adjustment magnitude (always 10%), the sunset window (always 15 years), the rebalancing logic (always PMR's five-step priority). The only knob is the initial withdrawal rate. Everything else flows from the paper.

This is the central design difference between the two strategies. Parametric guardrails give you four knobs and let you tune the strategy to your taste. Canonical GK gives you one knob, the initial rate, and the rest is the paper. The cost of the extra knobs is decision burden. The cost of the missing knobs is loss of control.

4. Implementing canonical GK in Bellavia

The canonical GK strategy is exposed in the calculator's withdrawal-strategy dropdown as "GK Guardrails (Original 2006)". It is a premium feature, as is the existing parametric guardrails strategy.

To run a canonical GK simulation, set:

• Initial portfolio — the dollar amount at the start of retirement
• Years — the planning horizon (the CPR-sunset rule kicks in for the last 15)
• Equity / bond / cash allocation — the initial allocation; the engine rebalances annually via the PMR
• Annual withdrawal — implies the initial withdrawal rate
• Equity index — pick the historical dataset (US equity, UK equity, custom)
• Real or nominal basis — for how the annual withdrawal is interpreted

That is the full input list. The four rules' thresholds and the PMR sourcing order are all fixed by the paper; the simulator does not expose them.

Two choices worth flagging

Cashflows are hidden under GK. The canonical Guyton-Klinger algorithm does not model lump-sum inflows or scheduled outflows. Adding a cashflow under GK would force the engine to make up rules the paper does not specify — which year of the WR/CPR/PR cascade does the cashflow modify? Should the rate computation include or exclude it? There is no defensible default. Rather than fabricate an answer, the calculator hides the "Add Other Cashflows" section as soon as you select GK. If you need cashflows, switch to the parametric guardrails strategy or one of the other withdrawal modes.

The allocation slider is an initial allocation, not a target. The PMR rebalances to the slider's weights at the start, but skim-to-cash and priority sourcing mean the allocation drifts intentionally over time. The calculator shows a small "GK uses its own rebalancing" notice next to the slider when you select GK so this is not a surprise. It is one of the few places where canonical GK's behaviour visibly diverges from what users expect from the other strategies.

5. Canonical GK in action

Take the same 1966 cohort and run it under canonical GK with the same 4% initial withdrawal rate. The chart below shows the year-by-year real withdrawal, with markers for each year that one of the rules fired:

Figure 2: Canonical GK on the 1966 cohort. The orange line is the real annual withdrawal. Red triangles mark CPR cuts (-10%), cyan triangles mark PR raises (+10%), and gray X marks WR freezes (the inflation bump was skipped). PR raises don't fire until the late 1980s — by then the portfolio has finally recovered enough that the (already-cut) withdrawal represents less than 80% of the initial 4% rate.

In this 30-year window, GK fires four CPR cuts, five PR raises, and three WR freezes. The cuts cluster in the 1970s: 1971, 1975, 1976, 1980 — exactly the years the parametric strategy was also snapping the withdrawal down. The mechanism is different: GK applies fixed 10% multiplicative cuts whereas the parametric strategy snaps to a fixed 5% rate. But the direction is the same.

The interesting feature of the GK trajectory is what happens between the trigger years. After a CPR cut, the new lower withdrawal carries forward (with the WR's freeze rule applied as needed). The 10% cut is permanent unless a later PR raise reverses it. This is the "cascading" nature of GK: one cut makes the next cut more likely if the portfolio keeps falling, because each cut shrinks the absolute withdrawal but the rate computation uses the post-withdrawal portfolio.

Compare this to parametric guardrails, which have no memory between years. The parametric strategy's withdrawal is recomputed from the current portfolio each year. If the portfolio recovers, the parametric strategy returns immediately to the target dollar amount. Canonical GK has to wait for the PR to fire which requires the rate to drop below 80% of initial, a much steeper recovery threshold.

6. Comparing the two

Here is the side-by-side, summarised:

Plotting both strategies on the same 1966 cohort makes the differences more clearly visible:

Figure 3: Both strategies on the 1966 cohort, same 4% initial withdrawal rate, $1M portfolio, 60/40 allocation. Parametric (green) changes responsively; canonical GK (orange) steps down through CPR cuts and waits for PR raises to recover. The dotted blue line is the constant $40K reference.

The parametric strategy is more responsive on the upside. When the portfolio recovers in the early 1980s, parametric's withdrawal climbs back faster than GK's, because GK can only raise via the PR's 10% step function and that requires a deep WR drop. The parametric strategy is also more aggressive on the downside. Its 5% upper bound bites harder than GK's 4.8% effective trigger (4% × 1.2), so cuts come earlier.

Across all cohorts

If we now zoom out from 1966 to every 30-year cohort in the historical dataset (91 cohorts spanning 1901-1990 entry years), running both strategies with the same 4% initial WR:

Figure 4: Total real dollars withdrawn over 30 years, by cohort start year. Parametric (green) and canonical GK (orange) both adapt; neither fails in any historical cohort at this withdrawal rate. The total real dollars differ cohort by cohort, with GK lagging in the worst stagflation cohorts and matching or exceeding parametric in the recovery decades.

Two patterns stand out:

• Both strategies survive every historical cohort. Constant-$40K would have failed in the worst stagflation cohorts. The point of the adaptive rules is to push the survival rate to 100% by trading lifestyle stability for portfolio durability. Both strategies do this, by different mechanics.

• The two strategies are not interchangeable on total dollars. In the 1965-1975 cluster — the worst entry windows in the dataset — parametric outperforms GK on total real dollars withdrawn. In the recovery cohorts of the late 1970s and 1980s, the gap closes and sometimes reverses. The ordering depends on the era.

A summary scorecard

Pulling together the same cohorts into a single table:

Five things this table says:

1. Parametric leaves more on the table on average. The mean ending portfolio in real dollars is about 6% higher under parametric. The mechanism is that parametric's snap-to-bound caps spending early and aggressively, especially in the strong-return cohorts where the upper guardrail rarely does and most years take the modest base $40K. GK is more willing to ride the PR (Prosperity Rule) up in those cohorts, so it spends more along the way and ends with less.

2. At the median, the ordering reverses. GK's median ending portfolio is slightly higher. The mean-vs-median gap reflects the asymmetric distribution of historical returns. A few exceptionally strong cohorts pull parametric's mean up because parametric leaves more upside unspent, but the typical cohort sees GK preserve slightly more capital.

3. GK is meaningfully more volatile. The average within-cohort standard deviation of real withdrawals is about $10,900 under GK vs $8,500 under parametric. This is roughly 29% more spending variability. The 10% step changes from CPR cuts and PR raises produce visible jumps; the parametric strategy's snap-to-bound is comparatively smoother.

4. GK's worst single year is better than parametric's worst single year ($17,784 vs $16,102). This is the surprise. Parametric's return-to-5% logic asymptotes as the portfolio shrinks, 5% of it shrinks too, with no floor. GK's 10% multiplicative cuts compound but more slowly than the asymptotic decay of "5% of a perpetually shrinking portfolio." In the worst historical cohort, GK's CPR cuts hit the floor of GK's 15-year sunset before parametric's snap-to-5% reached its lowest point.

5. On welfare-equivalent grounds, the two strategies are essentially tied. The certainty-equivalent annual consumption at γ=2 (a standard, modestly risk-averse utility) is $41,215 under parametric and $44,411 under GK — a 7.8% gap in GK's favour. The small difference is the net of GK's slightly higher mean withdrawals and its noticeably higher volatility. At gamma=2 these almost cancel.

That last finding, that the two strategies are nearly indistinguishable on welfare grounds, is the reason for the deeper question that Tharp & Fitzpatrick's 2024 critique posed: if neither strategy delivers appreciably better welfare, what should we use instead? That question will get its own follow-up post.

7. Conclusions

Two implementations of the same idea, that a retiree's spending should respond to the portfolio's behaviour, reach the same destination by different paths.

Choose parametric guardrails when:

• You want explicit control over the worst-case withdrawal rate. The upper bound is the cap on how much spending will fall as a fraction of the remaining portfolio. You set it.
• You want the strategy to recover quickly when the portfolio rebounds. The parametric strategy snaps back to the target dollar amount as soon as the current rate falls inside the bounds.
• You don't want to worry about a 15-year sunset on the cut rule, or about the PMR's asset-sourcing logic. Parametric treats the portfolio as a single pool.

Choose canonical GK 2006 when:

• You want a paper-faithful, fully-specified algorithm with three decades of practitioner literature behind it. GK is the most-cited dynamic-withdrawal strategy in retirement-planning research.
• You don't want to choose the band widths or adjustment magnitude yourself. GK fixes them, and the fixings are defensible — they are the values the original paper validated against historical data.
• You want the PMR sourcing logic. The "sell winners first" priority and the equity-skip sub-rule are non-trivial protections against selling into a drawdown. None of the other strategies in Bellavia model this.

Both strategies share an important caveat: neither is autopilot. Both demand annual review. The thing the strategies cannot do is tell you when your underlying assumptions need to change. A strategy decides the spending level given those assumptions. It does not decide the assumptions. That review remains the retiree's (or the advisor's) job.