不同評估績效的長短顯著地影響基金的經營策略，相較於強調穩健經營的退休基金而言，此因素是否亦影響退休基金的運作，本研究嘗試應用隨機控制理論，將投資績效的時間因素納入決策考量，以隨機微分方程式描述退休基金資產和應計負債的動態隨機行為，以多期基金規劃的觀點，探討時間因素與最適策略之關連性。本研究應用Brennan、Schwartz與Lagnado（1997）的結果至負債導向的退休基金管理，建構多期資產負債管理模型，退休基金持有資產將分類為風險性的股票投資組合、長期債券和短期票券，並考量投資標的短期利率與長期利率之隨機性質，將基金提撥與資產配置視為可調節因子，給定風險評估測度，於不設定投資限制下計算各期最適投資比例及基金提撥；本研究並以私人退休金個案進行模擬分析，結果顯示此基金未來10年之最適提撥率介於4.2﹪與5.1﹪，就不同評估期限而言，5年評估期之提撥率於初期高於10年評估期，基金比率η=0.75之提撥率低於η=1；5年評估期之基金交易行為較10年期明顯劇烈，基金比率較低時，其交易變化程度較小，不同評估年限與基金比率將同時影響退休基金之最適提撥與投資策略。 / Distinct time horizons in measuring investment perfomance significantly influence the financial planning for the money managers. In this study, we explore this issue concerning the pension fund management that has focused on the asset and liability management to meet its future obligations. A stochastic control model is formulated in a continuous-time framework to obtain the closed form solution for optimal strategy. The time variation in expected returns introduced in Brennan, Schwartz and Lagnado（1997）is adopted in obtaining the optimal strategy using plausible future plan’s normal costs and accrued liabilities under distinct time horizons. Based on the proposed performance measurement, the optimal funding schedule and portfolio selections are determined dynamically without trading restrictions. A private pension scheme is selected and analyzed for numerical illustration. It shows that the optimal contribution rates are between 4.2﹪and 5.1﹪for this specific case. Comparing the funding schedules for distinct time horizons, we find that the contribution rates under 5-year period are higher than those under 10-year period in the beginning. The contribution rates given funding ratio at 75﹪are lower than those given at 100﹪. While the optimal trading behaviors of the pension fund managers for 5-year period are significant volatile than those for 10-year period. Their optimal trading behaviors have exhibited a reduced volatility under the lower funding ratios. The case study indicates that the distinct time horizon and the funding ratio play crucial roles in decision-making process for pension fund management.

本文利用隨機控制理論，延續Chang et al. ( 2002 )，採用總和精算成本法，考慮提撥率風險( Haberman and Sung ( 1994 ) )極小的情況下，推導確定給付退休基金之最適提撥與資產配置策略封閉解，資產配置部分考慮股票市場投資組合、永續債券、現金三種部位。 套用公務人員退撫基金第四次精算報告之數據，透過Matlab重覆模擬1,000次，數值結果如下： 1.正常成本與提撥金額呈遞增趨勢，且兩數據差距甚小，符合風險評估函數所設定之提撥率風險極小化的要求。十年控制期間中，正常成本成長5.32倍，從1.03億增加至5.49億；提撥金額成長16.65倍，從0.33億增加至5.56億。275期以前正常成本大於提撥；275之後提撥大於正常成本。 2.初期提撥金額小於給付金額，且投資報酬不足以彌補其差額，因此造成基金規模縮小，但由於提撥金額成長速率大於給付支出，使得基金規模下降程度趨緩，隨後開始穩定成長。十年控制期間中，基金規模從起始的1,000億下降至840億，再上升至約1,314億。 3.股票與債券之持有或放空的部位越多，基金報酬率波動越大，基金規模越大時，可承擔風險的容量增加，因此傾向高風險投資；基金規模越小時，風險承受度變小，所以投資策略反而趨向保守。股票最多持有99.18%、放空90%；債券最多持有293.5%、放空140.14%。 / In this study, we continue using the model of Chang et al. ( 2002 ), which is based on stochastic control theory to study the dynamic funding policy and investment strategy for defined benefit pension plans. The model includes three investable assets: stock market portfolio, consol bond, and cash. We apply “Aggregate Actuarial Cost Method,” so only the contribution rate risk proposed in Haberman and Sung ( 1994 ) is considered when measuring the performance. In addition, we analyzed the data from Taiwan Public Employees Retirement System (Tai-PERS) investigate the optimal contribution and asset allocation through the proposed model and arrived at the following conclusion: 1.The trend of increasing normal cost and contribution as well as the small disparity tally with the requirement of minimum contribution risk as defined in the loss function. 2.In the beginning, the return of investment and contribution are insufficient to cover the benefit payment, causing the fund level to shrink; but as the rate of contribution increases over time and surpasses the benefit payments, the fund level will cease to shrink, and start to grow gradually. 3.There is a positive correlation between the fund level and the risk of investment. In other words, the larger the size of the fund level, the higher the possibility of holding or short selling risky assets.