Are Bonds Really Safer? Rethinking the Role of Fixed Interest in Long-Term Portfolios
For decades, financial advisers and investment textbooks have preached the same basic principle: as you get older, shift from equities to bonds. Reduce risk. Preserve capital. Secure retirement income. Bonds, it is assumed, are the ‘safe’ part of a portfolio.
The logic seems intuitive: equities may deliver growth, but bonds provide safety. Volatility is reduced, capital is preserved, and income becomes more predictable. But what if that entire framing is wrong? What if bonds are not merely less rewarding than equities, but genuinely riskier?
A landmark 2025 paper by Scott Cederburg, Aizhan Anarkulova and Michael S. O’Doherty, Beyond the Status Quo: A Critical Assessment of Lifecycle Investment Advice, launches an ambitious empirical assault on conventional asset allocation wisdom. Drawing on return data from 39 developed countries over more than 130 years, the authors simulate the financial trajectories of one million hypothetical couples saving for retirement, and their results are radical: portfolios invested entirely in equities—specifically, 33% domestic and 67% international stocks—outperform traditional stock-bond blends—products like target-date funds (TDFs)—on virtually every meaningful metric. More controversially, the research suggests that equities are not just superior in terms of long-term return potential, but are also more reliable, with fewer extreme negative outcomes than bond-heavy portfolios.
In other words, the asset class traditionally viewed as the ‘risk reducer’ may actually increase the odds of failure in retirement. This flips the conventional wisdom on its head—and demands scrutiny.
What the Research Actually Tested: Lifecycle Simulations on a Global Scale
Cederburg et al. (2025) develop a comprehensive lifecycle simulation framework to evaluate retirement outcomes for investor households. Each simulation models couples who begin saving at age 25, contribute 10% of their income annually until retirement at age 65, and then withdraw 4% of their accumulated savings per year (adjusted for inflation) throughout retirement. It is assumed that the household has a longevity of 95 years. The analysis spans 39 developed countries, drawing on asset return data dating back to the 1890s. By randomly selecting 10-year return blocks (called block bootstrap), the simulations incorporate a wide range of historical economic environments—including wars, inflationary episodes, recessions, and extended market downturns.
Crucially, the investment universe is simple and market-cap weighted: domestic equities (relative to each developed country studied), international equities, short-term government bills, and long-term government bonds. Corporate bonds, real assets, factor tilts, and inflation-linked securities are not included. This limitation is intentional—it ensures historical breadth and consistency across countries—but it also constrains the generalisability of the findings, as discussed later. The base case finding mentioned at the start of this blog assumes a fixed savings rate, a predetermined retirement age, and a static asset allocation.
However, dynamic optimisation is common in the lifecycle asset allocation literature—frequently allowing for variable savings rates, flexible retirement timing, age-dependent asset allocations, or responsiveness to market conditions. The authors test on dynamic optimisation from multiple perspectives but also note the methodology’s shortcomings: These models typically rely on simplifying assumptions, such as independent and identically distributed (IID) returns or linear relationships between expected returns and state variables. This is why the authors use block bootstrap in their own research—to make their findings theoretically more robust and more reflective of past historical economic environments.
To test time-varying investment allocations conditional on household age, the authors implement a computationally intensive iterative optimisation technique that determines the optimal asset mix at each age, whilst grouping early and late life stages (ages 25–29 and 90+) to mitigate simulation error. Again, it is assumed that the household has a longevity of 95 years.
The resulting strategy is remarkably similar to the optimal fixed-weight allocation: on average, households remain 99% invested in equities across the lifecycle. At age 65, the allocation temporarily shifts to 26% domestic stocks, 47% international stocks, and 27% short-term bills—a modest adjustment in response to sequence-of-returns risk under the 4% withdrawal rule. By age 68, however, the equity allocation once again exceeds 90%, and bonds are never meaningfully introduced. The average asset allocation split is 31% domestic stocks and 69% international stocks.
Importantly, a couple following this age-based allocation would need to save 9.9% of their income to achieve the same expected retirement utility as a 10.0% savings rate under the fixed-weight strategy. This indicates that dynamic allocation offers little economic advantage beyond what is achievable through an appropriately chosen static approach.
Other scenarios that were tested include, but are not limited to:
Income and age-based contribution rates
Time-varying investment allocations conditional on the market state
For the former, the couple still chooses an allocation of 33% in domestic stocks, 67% in international stocks, 0% in bonds and 0% in bills. This modification to the static design with a constant savings rate does not affect the optimal strategy.
For the latter, the authors’ dynamic model provides households with monthly information on valuation quintiles, from cheapest (price-dividend < 20) to most expensive (price-dividend > 50). When the domestic price-dividend ratio is low, investors weight domestic stocks heavily at 65% and allocate the remaining 35% to international stocks. In the middle three quintiles, the allocations are similar to the unconditional, fixed-weight optimal strategy. In the quintile with the highest price-dividend ratios, the couples reduce the domestic stock allocation to 16%, increase international stock exposure to 75%, and invest 9% in bonds. Thus, when domestic stock prices are very high, the couples optimally allocate a small portion of their wealth to bonds. For comparison, to achieve the same expected retirement utility as the couple saving 10.0% with the optimal fixed-weight strategy, a couple using the conditional market state strategy saves 9.7%.
Why the Case Against Bonds is Stronger Than It Seems
The study’s findings rest on a fundamental insight: bonds may offer nominal stability but pose substantial real risk. This is particularly acute over long horizons and during retirement, when inflation and longevity combine to compound the downside.
Consider this:
Over a 10-year horizon, 30% of bond simulations produce negative real returns, compared to 22% for domestic equities and 18% for international equities (Cederburg et al. 2025).
During periods of high inflation, such as the 1970s or more recently 2021–22, bonds suffered massive real losses: US 10-year Treasuries lost around 30% in real terms over just two years (Credit Suisse 2022).
Cederburg et al. (2025) find that in deep left-tail (worst-case) outcomes, the optimal portfolio—made up of 33% domestic and 67% international equities—outperforms the TDF strategy in preserving capital.
In deep right-tail (best-case) outcomes, the optimal portfolio significantly outperforms the TDF strategy in accumulating capital—as would be expected.
On average, the optimal portfolio outperforms the TDF strategy at both preserving and accumulating capital.
The key behavioural point is that these risks are often invisible to investors looking at nominal return streams. Bonds may seem stable because they do not swing wildly month to month—but they can silently fail to deliver the purchasing power needed in retirement, especially if combined with longevity or rising living costs. The authors find that in portfolios modelled after common TDFs, the probability of running out of money during retirement under a 4% withdrawal rate is nearly 20%. In contrast, the fully equity-based optimal strategy has a ruin rate closer to 7%.
Inflation-Adjusted Retirement Consumption
One of the most practically relevant measures of retirement success is not portfolio value or volatility, but the level of inflation-adjusted consumption that a household can sustain throughout retirement. The authors find that the all-equity strategy delivers markedly better outcomes than traditional bond-inclusive portfolios. Table IV of the paper (above) reports average real replacement rates—that is, the ratio of inflation-adjusted retirement income to pre-retirement earnings—for a range of strategies. The optimal portfolio supports an average replacement rate of 1.24. A replacement rate of 1.24 means that, on average, a retiree following the optimal portfolio strategy would receive 124% of their last working year's income (adjusted for inflation) each year throughout retirement. This is not a one-time figure or cumulative total—it's a sustained annual level of real consumption.
More notably, around 55% of simulated households in this strategy achieve full or greater replacement. Even in the lower tail of the distribution, real consumption outcomes for the equity-only strategy surpass those of typical target-date funds, suggesting not only stronger average performance but also better protection against poor retirement income scenarios. This highlights a central insight of the paper: that equity portfolios—particularly those diversified globally—are not only capable of growing wealth, but are also more effective at sustaining real purchasing power in retirement.
Germany
As an interesting aside, another fascinating observation is the standard deviation of bond returns for Germany. This is extreme relative to the estimates for other sample countries and is attributable to the well-known period of hyperinflation in Germany in the 1920s. Excluding all German data from the model–which would make bonds appear less risky and thus more favourable to the hypothetical couples–does not change the optimal fixed-weight asset allocation policy (Anarkulova, Cederburg, and O’Doherty 2025).
TLDR;
The key outcome metrics evaluated in the study include:
Wealth accumulation at retirement
Probability of running out of money during retirement (the ‘ruin rate’)
Portfolio outcomes in the tails of the distribution (e.g. 1st-10th or 90th-99th percentile of wealth or spending)
Inflation-adjusted retirement consumption
Across all of these dimensions, the 100% equity strategy consistently outperforms. It generates higher terminal wealth, supports better consumption outcomes, and reduces the likelihood of retirement ruin—even in the most adverse simulated market conditions.
Intelligent Pushbacks: Key Limitations and Critical Considerations
Whilst Beyond the Status Quo is an ambitious and data-rich examination of lifecycle asset allocation, it is not immune from critical scrutiny. Several well-founded objections have been raised by academics, practitioners and commentators in the evidence-based investing community. These do not negate the paper’s findings but highlight areas where its assumptions, exclusions, and interpretations may limit its applicability to modern, real-world portfolios. The most important critiques concern inflation protection, current market conditions, diversification effectiveness, asset class representation, behavioural constraints, and the interpretation of correlation robustness within the study’s framework.
1. The Exclusion of Real-Return Bonds (Inflation-Linked Securities)
A notable omission from the simulation is the exclusion of inflation-linked bonds, such as US Treasury Inflation-Protected Securities (TIPS) or UK index-linked gilts. These instruments are specifically designed to protect investors from inflation erosion—one of the key risks highlighted in the paper’s critique of nominal bonds.
The absence of real-return bonds may overstate the riskiness of bonds (fixed income), particularly for retirees seeking to maintain purchasing power throughout retirement. Inflation-linked bonds provide a structurally different return profile from nominal bonds and are increasingly regarded as a foundational component in retirement income strategies. Although they are not without limitations—such as typically lower yields, liquidity constraints, and dependency on official inflation measures—they can play a critical role in hedging long-term inflation risk.
It remains an open question whether portfolios blending equities with inflation-linked bonds might deliver similar retirement outcomes to a 100% equity strategy, but with lower behavioural and volatility risk. By excluding these securities, the study narrows the potential range of defensible fixed income strategies and weakens the generalisability of its conclusions.
2. Elevated Equity Valuations and Forward-Looking Return Assumptions
Another significant limitation concerns the current market environment. At the time of writing, equity valuations—particularly in the United States—are historically elevated. As acknowledged by Cederburg in subsequent commentary, US stocks are currently within the highest valuation quintile (price-to-dividend ratio exceeding 50), a regime that was only sparsely represented in the historical data used to simulate returns.
This raises the question of whether the next several decades will resemble the average of the past 130 years. If equity returns are structurally lower due to valuation (mean) reversion, the relative advantage of all-equity portfolios could narrow. Additionally, if the margin of safety that equities historically offered over bonds diminishes, the case for including lower-return but less volatile assets becomes stronger.
Furthermore, as noted by RobertT in the Rational Reminder community, the study’s use of block bootstrapping from historical data likely underrepresents recent high-valuation, high-correlation periods. As such, the simulated outcomes may overstate both the diversification benefits of international equities and the long-term return potential of equity portfolios initiated in a high-valuation environment.
3. The Fragility of International Diversification
A foundational pillar of the study’s optimal portfolio is global diversification. The suggested 33% domestic and 67% international equity split rests on the historical benefit of relatively low correlation between domestic and foreign markets—averaging approximately 0.33–0.34 over the sample period. However, such figures may no longer reflect the structural reality of modern markets.
In recent decades, capital markets have become significantly more integrated. Cross-country equity correlations, particularly between the US and other developed markets, have often exceeded 0.8 (Credit Suisse 2022; Anarkulova et al. 2020). This elevated correlation reduces the marginal benefit of international diversification and may diminish the value of holding equities abroad, particularly during global downturns.
More problematically, the assumption of frictionless international investing is likely too optimistic. As highlighted by the Credit Suisse Yearbook (2022), the periods during which diversification is most needed—such as war, geopolitical instability or systemic crises—are precisely the moments when international investing becomes most difficult. Historical precedents such as World Wars I and II and the Great Depression saw capital controls, expropriation of foreign assets and restrictions on foreign ownership (Esteves and Khoudour-Castéras 2009). In such scenarios, the assumed benefits of global equity exposure may fail to materialise.
Thus, whilst international equities remain a key source of return diversification under normal conditions, their reliability as a crisis hedge is open to question. The assumption of constant and accessible diversification may understate the risks associated with global equity strategies.
4. A Narrow Representation of the Investable Universe
The asset universe used in the simulations is intentionally narrow: domestic equities, international equities, government bonds and short-term bills. Whilst this enhances data quality and historical depth, it excludes numerous asset classes that form part of a modern investor’s toolkit.
Notably absent are corporate and high-yield bonds, which can offer improved risk-adjusted returns relative to sovereign debt; inflation-sensitive real assets such as commodities, infrastructure, and real estate; and factor-based equity strategies (e.g. value, profitability, quality, momentum) which have demonstrated persistent long-term performance (Fama and French 2015; Novy-Marx 2013).
The absence of these assets limits the scope of the study’s conclusions. For example, the poor historical performance of nominal government bonds does not necessarily imply that all forms of fixed income are suboptimal. Likewise, the use of market-cap-weighted equity portfolios excludes the possibility that a diversified set of equity risk factors might reduce volatility or improve tail outcomes—without requiring a full bond allocation.
Consequently, whilst the findings are internally consistent within the tested universe, they should not be interpreted as broadly applicable to all equity or fixed interest strategies currently available.
5. Behavioural Considerations and the Limits of Theoretical Optimality
Even if the 100% equity portfolio is optimal in a strictly statistical or utility-maximising sense, its behavioural feasibility for most investors remains doubtful. Retirees, in particular, may find it psychologically untenable to hold an all-equity portfolio through major drawdowns.
Market crises—such as those in 2000–02, 2008–09 and 2020—can result in portfolio declines of 30–50%. For retirees living off portfolio withdrawals, this introduces severe sequence risk. More importantly, such declines often trigger emotional responses, leading to ill-timed de-risking, withdrawal adjustments, or full capitulation.
As Ben Felix and Dan Bortolotti note in Rational Reminder AMA #4 (2025), ‘a client who spends less when markets are down and more when they’re up is not likely to be happy’. Many investors do not merely seek maximum expected consumption—they also desire predictability, peace of mind and a stable standard of living.
For this reason, even if bonds do not improve long-term portfolio efficiency, they may still serve as behavioural anchors, helping investors to adhere to their strategy in times of stress. A portfolio that is theoretically optimal but behaviourally intolerable is unlikely to succeed in practice.
6. Panel D and the Inadequacy of the Correlation Analysis
In addressing the concern that international diversification may falter during high-correlation regimes, the authors introduce ‘Figure C.1. Conditional ruin probabilities’ and, specifically, panel D, which segments retirement periods based on the observed correlation between domestic and international equity returns. The results appear to show that the all-equity strategy outperforms across all correlation quintiles, with lower probabilities of running out of money even during periods of elevated correlation.
However, this analysis has been criticised for its lack of granularity and real-world relevance. As RobertT in the Rational Reminder community observes:
‘Just my personal opinion, but I do not think the paper fully addresses the correlation concerns. It would have been more instructive for Figure C.1.D to split the results by correlation coefficient e.g. 0.2, 0.4, 0.6, 0.8 (or more specifically: 0 to 0.2, 0.2 to 0.4, 0.4 to 0.6, 0.6 to 0.8, 0.8 to 1) rather than “low” to “high” quintiles.’
This is a salient point. By using historical quintiles, the highest correlation group may still average well below 0.8, which is now typical in developed markets (Credit Suisse 2022). Therefore, the analysis may understate the challenges posed by today’s high-correlation regime, and may not reflect the degradation of diversification benefits observed in modern portfolios.
A more robust approach would involve segmenting outcomes by actual correlation coefficients, allowing investors to better understand how the proposed strategy behaves under current and likely future conditions. Without such specificity, Panel D offers reassurance without sufficient empirical precision.
Final Thoughts: Bonds, Behaviour, and the Limits of Modelling
The research by Cederburg and his co-authors delivers a powerful and well-argued challenge to conventional wisdom about fixed interest. By showing that equities not only outperform over the long run but also offer more resilient outcomes in worst-case retirement scenarios, the paper repositions bonds as a behavioural crutch more than a financial necessity. And yet, that may be exactly the point.
Whilst the historical evidence tilts heavily toward equities, the practical reality is far more nuanced. Today’s valuations are elevated. Global markets are more closely correlated than ever. The benefits of international diversification may not materialise in systemic crises. Inflation-linked bonds, corporate debt, and factor-based equities all add further complexity—and opportunity.
Most importantly, investors are not robots. Retirement is not a Monte Carlo simulation. It is a lived experience, shaped by emotion, uncertainty, and often fear. The best strategy is not the one that performs best on paper, but the one that a retiree can stick with through the worst of times.
Bonds are not inherently flawed—but their role may need to be reframed. They are not there to enhance returns. They are there to help people sleep at night. The true challenge for investors and advisers is not simply to seek higher expected returns, but to construct portfolios that balance robust long-term outcomes with behavioural realism.
References
Anarkulova, Aizhan, Scott Cederburg, and Michael S. O’Doherty. 2025. Beyond the Status Quo: A Critical Assessment of Lifecycle Investment Advice. Working paper, University of Arizona.
Cederburg, Scott, Michael S. O’Doherty, and Richard Sias. 2020. ‘The safe withdrawal rate: Evidence from a broad sample of developed markets.’ Journal of Financial Economics 138 (3): 820–839.
Credit Suisse. 2022. Global Investment Returns Yearbook 2022. Zurich: Credit Suisse Research Institute.
Esteves, Rui Pedro, and David Khoudour-Castéras. 2009. ‘Remittances, capital flows and capital controls: Evidence from the Great Depression.’ Journal of Economic History 69 (1): 125–158.
Fama, Eugene F., and Kenneth R. French. 2015. ‘A five-factor asset pricing model.’ Journal of Financial Economics 116 (1): 1–22.
Novy-Marx, Robert. 2013. ‘The other side of value: The gross profitability premium.’ Journal of Financial Economics 108 (1): 1–28.
Rational Reminder. 2023. ‘Episode 284: Prof. Scott Cederburg – Challenging the status quo on lifecycle asset allocation.’ Rational Reminder Podcast, 7 December.
Rational Reminder. 2025. ‘Episode 349: AMA #4 – Retirement spending and investor behaviour.’ Rational Reminder Podcast, 20 March.
Rational Reminder. 2025. ‘Episode 350: Scott Cederburg – A critical assessment of lifecycle investment advice.’ Rational Reminder Podcast, 27 March.