Investment risk-taking and benefit adequacy under automatic balancing mechanism in the Japanese public pension system

Japan's public pension system, a pay-as-you-go system with substantial reserve funds, faces challenges due to declining birth rates, aging populations, and increasing longevity. Maintaining the system's long-term sustainability while ensuring adequate benefits is crucial, yet these goals are often conflicting. The 2004 introduction of an automatic balancing mechanism (ABM) aims to restore financial equilibrium through benefit adjustments based on solvency indicators (life expectancy, demographic ratios, and wage growth). The ABM's non-linear functions make analyzing the impact of market fluctuations complex. This study uses a stochastic simulation approach, leveraging Japan's official pension verification programs, to assess the risks to benefit levels (measured by RR) and financial stability (measured by RER) under different investment risk-taking levels in the reserve fund, controlled by the SW (percentage of investments in stocks). This contrasts with existing scenario-based predictions which do not fully capture the short-term fluctuations and risks inherent in a dynamic system with an ABM. The government's target replacement rate (at least 50%) and the reserve-to-expenditure ratio (RER = 1 at the end of a 95-year projection period) serve as key performance indicators. Previous studies have examined ABMs' effectiveness but often separately assess benefit adequacy and financial sustainability, and rarely incorporate the impact of varying investment strategies, particularly in the context of a large reserve fund invested in equities. This study addresses this gap by simultaneously assessing the downside risks of RR and RER under the ABM and exploring the effects of dynamic investment strategies.

Several studies have explored automatic balancing mechanisms (ABMs) in public pension systems and their impact on benefit adequacy and financial sustainability. Vidal-Meliá et al. (2009) investigated the introduction of an ABM in Spain, finding that it created a credible institutional framework and minimized the system's use as an electoral tool. Boado-Penas et al. (2020) assessed a pay-as-you-go system with an ABM, identifying trade-offs between sustainability and adequacy based on different risk-taking strategies for the investment fund. Fujisawa and Li (2012) examined the impact of the Japanese ABM on pension benefits using a stochastic simulation focusing on mortality and fertility rate fluctuations. However, prior research often lacked a comprehensive stochastic simulation model incorporating the complexity of the Japanese ABM and the significant role of reserve fund investment risk-taking, which is the main focus of this study. Unlike previous studies that varied contribution rates, pension ages, or indexation rates, this study uniquely focuses on the impact of equity risk-taking (SW), the only controllable parameter for the government under the existing ABM in Japan. This study builds upon the work of Kitamura et al. (2006), MHLW (2020b), and Nakashima and Kitamura (2021) by using a more detailed stochastic simulation of the Japanese public pension system that explicitly models the ABM's non-linear impact on benefits, and incorporates the stochastic nature of various macroeconomic factors.

The study employs a stochastic simulation model based on the Japanese Ministry of Health, Labour and Welfare's (MHLW) official pension verification program. This program projects the long-term financial condition of the public pension system over approximately 100 years, incorporating assumptions about demographics, social factors, and economic conditions (including inflation, wage growth, and investment returns). The model simulates the impact of different stock weights (SW) in the reserve fund investments on the replacement rate (RR) and the reserve-to-expenditure ratio (RER). The simulation accounts for the automatic balancing mechanism (ABM), which adjusts pension benefits based on the RER. The ABM's modified indexation rate (m) is calculated based on inflation (θ), wage growth (φ), and the growth rate of insured persons. The model uses the MHLW's methodology for projecting insured persons, pensioners, pension benefits, premium revenue, total expenditure, and reserve funds. The simulation focuses on economic fluctuations by using the medium variant of population projections and two government economic cases (Case 3 and Case 5 from MHLW (2019b)). Case 5, representing a pessimistic scenario, is the primary focus to assess downside risk. The stochastic variables (inflation rate, wage growth rate, and investment returns) are assumed to follow a normal distribution. Parameters like mean, standard deviation, and correlation are based on the GPIF (2020a) data, supplemented by estimations using 25 years of historical data (1993–2017). Five SW levels (0%, 25%, 50%, 75%, 100%) are considered, with equal weighting for domestic and foreign assets within each SW level. The model projects the RR and RER over the period from 2019 to 2115, with 9000 Monte Carlo simulations run for each SW and economic assumption. The study also explores dynamic investment strategies, portfolio insurance and contrarian strategies, adjusting the SW based on RER levels at specific intervals (2040, 2060, 2080, 2100). The conditions for changing SW are defined based on pre-set RER thresholds, simulating both strategies. The present value of the government's obligation to compensate for negative RER (reserve fund depletion) is calculated using four discount rates: average 10-year JGB yield, expected return on domestic bonds, inflation rate, and expected return on investments, with both cashflow-based and balance-based measures used to assess this cost.

The study's key findings demonstrate the significant downside risks of both RR and RER, even with the ABM in place. Under the pessimistic economic scenario (economic assumption 5), the 10th percentile values of RR consistently fall below 40% by around 2060, and further decrease with increasing SW, as the ABM continues to reduce benefit levels in some simulation paths. High SW scenarios (75% and 100%) show similar 90th percentile values because of the upper limit imposed by the ABM on benefit indexation. For the RER, the 10th percentile values can fall below zero under all SW levels, indicating a potential for reserve fund depletion. Even when the RER recovers to one at the end of the simulation period due to benefit reductions, the prolonged effect of the ABM on benefit levels causes the 10th percentile of the RR to remain persistently low. Under the standard economic scenario (economic assumption 3), the overall levels of RR and RER are higher but maintain similar trends regarding SW. Joint distributions of RR and RER reveal that while moderate SWs (50% and 75%) yield the highest median values, lower SWs provide greater stability regarding VaR and CVaR. The analysis of a specific year (2075) shows a clear relationship between SW and the joint distribution of RR and RER; moderate SWs offer the best 50th percentile values, while lower SWs provide greater stability in VaR and CVaR. This pattern is consistent across 2050, 2075, 2100, and 2115. The present value of the government's obligation for reserve fund depletion is significant, ranging from JPY 9.5 to 65.3 trillion, depending on the calculation method and discount rate used. Sensitivity analysis reveals that investment return volatility primarily drives the fluctuations in RR and RER, significantly more than inflation or wage growth. Doubling volatility increases risk, while halving it reduces risk; however, the effects are not uniform across all percentiles. Allocating solely to domestic or foreign assets is riskier than a diversified portfolio. Dynamic investment strategies show that portfolio insurance effectively limits downside risks but at the cost of lower median RR and RER values compared to the constant mix strategy. The contrarian strategy increases median RR and RER but offers less downside risk protection.

This study's findings highlight the complexity of managing risk and achieving both sustainability and benefit adequacy in Japan's public pension system. The results challenge the government's current scenario-based prediction model, demonstrating that the replacement rate may be considerably lower than predicted under various scenarios. The optimal level of stock weight (SW) represents a critical policy decision, requiring a trade-off between maximizing median RR and RER and minimizing their downside risks. The study's results suggest that a moderate SW (around 50%) may strike a reasonable balance between these conflicting objectives, offering a relatively stable and reasonably high median RR and RER, while minimizing the chances of extreme outcomes such as reserve fund depletion. The significant potential government burden associated with reserve fund depletion further underscores the need for a cautious approach to investment strategy. The impact of investment volatility underlines the importance of considering a range of economic scenarios and incorporating risk management strategies into the long-term planning of the pension system. The analysis of dynamic investment strategies provides insights into potential ways to mitigate downside risks, but each strategy has its own trade-offs. Portfolio insurance shows promise in enhancing downside risk protection, while the contrarian strategy boosts median performance but may not provide as much protection. Future research should explore more sophisticated dynamic investment strategies and their impact on RR and RER, and refine the model to incorporate potential defaults on government bonds.

This study offers valuable insights into the risk-return trade-offs associated with different investment strategies in Japan's public pension system. The findings suggest a need for a more nuanced approach to risk management, potentially involving a moderate level of risk-taking in the reserve fund and consideration of dynamic strategies to mitigate downside risks. The considerable potential cost of reserve fund depletion highlights the importance of considering government borrowing as a potential measure to ensure the system's long-term viability. Future research should examine the interaction between the pension system and the wider macroeconomic environment, explore more complex dynamic investment strategies, and consider the implications of incorporating alternative asset classes. Incorporating the insights of behavioral finance into the decision-making process would also be a valuable area for future research.

The study relies on several key assumptions, including the normality of stochastic variables and the absence of government bond defaults. The assumed constant volatilities and correlation structure may also differ from real-world conditions. The focus on only two polar cases (domestic-only and foreign-only allocation) in the sensitivity analysis limits the conclusions on optimal asset allocation. Finally, the model does not explicitly account for certain factors that could influence the dynamics of the system, such as changes in government policy or unexpected economic shocks.