高鐵在建造過程中的巨大超支和延誤曾引起社會嘩然。港鐵為文過飾非邀請了兩位國際專家作了研究報告。在2014年7月港鐵發表「獨立董事委員會就高鐵香港段調查之第一份報告」，隨後在10月發表第二份報告，及在12月發表專家報告。
專家指出政府和港鐵的管理方法落後於國際慣常做法。專家對預算控制提了3個意見（註1）。筆者為此寫了兩篇文篇，分別為：高鐵超支之「分級備用資金」和【高鐵超支延誤】甚麼是參考組別預測？
本文完成第三部份，「概率成本指標」（中文版），或「備用資金的透支」（專家報名原文 contingency draw down ）（註2）。備用資金的設立是為了應付承建商的申索。筆者在張炳良在2014年5月4日的疑中留情事發之前，從高鐵的申索開支中發覺出了問題，並於2014年3月26日撰文，《張炳良隱瞞高鐵總站的申索問題》及在3月27日提交立法會，《對高鐵總站的申索的意見書》。筆者的結論是，「由此看來，當日蘋果日報確是捉錯用神，當中既非南昌站，也非工程進度，也不是單純的申索超出預算，而是西九總站的選址和其地盤勘測為何出現嚴重偏差」。筆者當時運用的手法也是類似於備用資金的透支。備用資金還包括財力保證金、履約保證金和進度保證金的分期支出（including retention, bonding and guarantees）。
概率成本指標
專家對港鐵或港府慣用的基本工程開支外加一個固定的比成作備用資金及對其透支缺乏敏感度提出了補充意見（註3）。專家發覺，工程超支的危機最容易反映在備用資金的支出（註4）。
專家翻查資料，發現備用金開支從2011年至2014年，平均延誤4%發放，但其支出却比實際進度超出23%，遠遠高於行業標準（註5）。
專家指出，備金資金的透支速度可以反映出：
一，估算未來的總超支；
二，與一些過去的相類型項目比較出其超支的概率分布。
總的來說，專家認為政府（港鐵）的現行工程預算不合格（註6）。
結語
報告歸報告，專家的意見方法其實不是新鮮事物，早已記戴在各類文獻中，港鐵的項目總監和港府的高級工程師冇可能不知道。一位退休政府工程師有點動氣地如此答覆筆者，「既然好的方法已經存在，他會問：現在的問題是否方法問題，還是其他問題？」
備註
註1
獨立專家的建議
2.37為加強管理與提取備用資金有關的成本風險，獨立專家建議採用以概率為基礎的關鍵績效指標，以監察高鐵項目在預算金額內完成的機會。
預算控制
1.33獨立專家審視了高鐵香港段目前的預算控制安排。獨立專家認為，成本超支的根本原因，是最初的預算安排過緊。成本超支歸因於兩個因素：（1）工程延誤及（2）香港建築業過熱導致市場供應的壓力。為應對日後項目的成本超支風險，獨立專家提出了三項建議，即港鐵公司採用：（1）概率成本指標；（2）利用參考組別預測來制定合理的預算及備用資金，以增補現行的成本計算的做法；及（3）在「服務經營模式」下為項目設立分級備用資金。
概率分析的步聚
1、列出各種欲考慮的不確定因素。需要注意的是，所選取的幾個不確定因素應是互相獨立的。
2、設想各各不確定因素可能發生的情況，即其數值發生變化的幾種情況。
3、分別確定各種可能發生情況產生的可能性，即概率。各不確定因素的各種可能發生情況出現的概率之和必須等於1。
4、計算目標值的期望值
6.4 In order for the MTRCL budget control system to become better capable of quickly detecting and hedging against such potential cost overruns, the Independent Experts recommend:
i. The use of contingency draw down as a leading indicator to detect escalation earlier in the project cycle;
ii. The use of reference class forecasting to establish appropriate contingencies for highly complex and uncertain projects; and
iii. The establishment of a tiered contingency fund, in particular for projects funded by Government under the “concession approach.”
註二
6.4 In order for the MTRCL budget control system to become better capable of quickly detecting and hedging against such potential cost overruns, the Independent Experts recommend:
i. The use of contingency draw down as a leading indicator to detect escalation earlier in the project cycle;
ii. The use of reference class forecasting to establish appropriate contingencies for highly complex and uncertain projects; and
iii. The establishment of a tiered contingency fund, in particular for projects funded by Government under the “concession approach.”
註三
6.2 The cost control processes currently in place for MTRCL projects track cost development against the contract baseline (i.e. the Initial Contract Control Total). The overall cost risk position is regularly estimated by the PjT. The risk and contingency status are reported monthly to the GM, PjD and the Project Control Group (PCG). If a contract exceeds its defined cost control total a change request is made to the PCG. Approved change requests are funded from the available project contingencies. MTRCL has weekly PCG meetings, which ensure a fast process of approving or rejecting claims and change requests.
6.3 MTRCL’s system of estimating and controlling cost has been used with success in previous MTRCL projects. Only one of six past projects exceeded their budget: the Lantau Airport Railway by 6%.
6.5 The current process of cost controls tracks the cost implications of events that impact the Project schedule. The monthly Project reports include an overview of additional funding needs and project by project reports on the estimated final project cost, contingency analysis and cost to complete outlook. The project dashboard compares a project’s actual and planned payment schedule. Currently, no traffic light indicator is being used to report on Project cost.
註四
6.9 As discussed in Part II, the Independent Experts find that the key cost risks for the XRL are best reflected in the draw down of contingencies. The contingency funds are used to pay for the extension to the Project programme of moving the opening date from 2015 to 2017 and unforeseen delays to finish the whole of works. The Independent Experts recommend that this cost risk is closely monitored by including in the project summary the monthly
註五
6.6 The Independent Experts reviewed past XRL progress reports for the period from November 2011 to July 2014. The Independent Experts found that the actual monthly payments consistently lagged behind the planned payment schedule by an average of 4%.
6.7 However, actual payments were consistently ahead of the actual progress by on average 23%. This is unusual compared to industry practice, but members of Procurement explained in interviews that it is standard MTRCL practice and that it is done to ensure cash flow to contractors. The key benefit was described to the Independent Experts as increasing contract security, including retention, bonding and guarantees. This practice might lead to the loss of advance payments in the case of insolvency of a contractor
註六
6.11 For example, a trending indicator of the contingency draw down could indicate (1) a conservative estimate of the total outturn cost, provided contingencies are drawn down at a constant rate,
(2) the probability of staying within the currently forecasted HK$71.5 billion outturn cost, i.e. a P-value based on actual outturn cost and actual cost overruns in a reference group of similar past projects
6.12 Additionally, project cost reporting should make it easier to establish whether a project is likely to overrun its budget and if so by how much. This could be achieved by including a cost performance index similar to the schedule performance index that is currently part of the progress reports. Other information to report could be the overall cost contingency position, major changes to the contingency position and a forecast of total outturn cost.
註七
6.10 The current practice of constructing a most likely, a pessimistic and an optimistic scenario for the cost position of projects follows the logic of earned value management systems, which make the assumption that current variations from the cost baseline are of a temporary nature. However, experience from large, complex projects, like XRL, show that cost variations on such projects are often systemic, caused by optimism in initial budgets. Thus scenarios for the most likely, optimistic, and pessimistic outturn cost should also incorporate a scenario where the actual historical volume of claims and variations is projected into the future.