《建设项目风险管理》课程教学参考资料：discussion of SHEN paper

DISCUSSIONS AND CLOSURES Discussion of "Risk assessment for where Cov(a,B)represents the covariance between a and B such Construction Joint Ventures in China that Cov(a,B)=[(a-a)(β-β)/54], and a andβ represent the mean values of a and B respectiv by L.Y. Shen, George W. C. Wu, In addition, we have to assume that a and B are independe and Catherine S.K. Ng hat is, how the respondents perceive the probability that risk January/February 2001, Vol. 127, No. l, pp. 76-81 occurrence is not affected by its corresponding perceived degree of risk impact. This means that Cov(a, B)=0, implying E(aB) Raymond Y. C. TSe E(a)E(B). Following this way, the correct calculation of the Department of Building and Real Estate, Hong Kong Polytechnic Uni- ndex score should be versity, Hung Hom, Kowloon, Hong Kong RS(k)=∑a/54p/54 China is the largest developing country in the world. The country has attracted worldwide attention. Although China has a protected Thus. the risk significance index would be underestimated construction projects in many ways Above al, f ate in Chinas using Eq(I). The authors should first separately calculate the the government must make efforts to create a better investment would be encountered if a high degree of loss tends tod problem ventures play an important role in the construction industry. Thus, is that a and B may not be independent. However, the environment and strengthen its cooperation with multinational ated with a low probability of occurrence. In this case, we assume companies to ensure the sound development of the national RS=a(B)B. Therefore,dRS/dB=α+β(daldβ). This suggests of its foreign investment. While the government further opens the or negative, depending on the value of daldB. If Cov(a, B) private sector to foreign investors for the establishment of joint #0, it is not difficult to estimate Cov(a, B) using the existing ventures, the associated risk environment has to be fully recog- data. With a sufficiently large sample, the author may examine nized Cov(o, B)for each risk item. One would expect that Cov(a, B)is A good estimate of risk requires a careful analysis of business negative nditions and their potential impacts. However, it is difficult to Because the variables take a discrete value, such that high quantify the perceived risk of construction joint ventures. The =1.0. average=0.5. and low=0.1. the domain of the risk signifi authors have made an interesting contribution in the area of risk cance index should lie in the range [0.01, 1). This means that th estimating by presenting a simple scoring method for analysis of significance index that has a value exceeding 0.25 can be classi- the perceived probability of risk occurrence and degree of risk fied as"high risk. "Alternatively "low risk"is indicated if the impact. Although the study is a good solution to the problem index is less than 0. 25. In this case, the t test can be applied to the hand,the method presented is not without problems. The dis- null hypothesis RS=0.25. Based on the t test, the authors are cusser feels several procedural questions arise from the paper. advised to categorize the risk significance index into three levels Let us begin with a brief examination of how the risk signif low risk for RS0.2 contains two attributes: (1)the probability level of the risk occur- Alternatively, it is more appropriate to use frequency analysis rence(a)and(2)the degree of impact or the level of loss if the for such study. The authors should indicate the respective fre risk occurs(B). Thus, in the authors paper, for each respondent quency (in percentage) of the high, average, and low risk of each the risk significance is given by risk item The frequency data shown in Table I shows that the score for factor). In Table RS(k)=2 sy/ Table 1. Frequency Analys In this case, with an impact value (B)and a probability of occur rence(o), the expected value of the impact is actually E(aB) Frequency (%) (% Expected value =aE(s). However, the problem is that both a and p are"per-High ceived"values. Using this approach, the expected value of the Average product aB should be 0.004 E(αB)=E(a)E(B)+Cov(a,β) Score 0.214 194/JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT/ MARCHIAPRIL 2002

Discussion of ‘‘Risk Assessment for Construction Joint Ventures in China’’ by L. Y. Shen, George W. C. Wu, and Catherine S. K. Ng January/February 2001, Vol. 127, No. 1, pp. 76–81. Raymond Y. C. Tse Department of Building and Real Estate, Hong Kong Polytechnic Uni￾versity, Hung Hom, Kowloon, Hong Kong China is the largest developing country in the world. The country has attracted worldwide attention. Although China has a protected market, global institutional investors can participate in China’s construction projects in many ways. Above all, Sino-foreign joint ventures play an important role in the construction industry. Thus, the government must make efforts to create a better investment environment and strengthen its cooperation with multinational companies to ensure the sound development of the national economy. This is also important in order to ensure stable growth of its foreign investment. While the government further opens the private sector to foreign investors for the establishment of joint ventures, the associated risk environment has to be fully recog￾nized. A good estimate of risk requires a careful analysis of business conditions and their potential impacts. However, it is difficult to quantify the perceived risk of construction joint ventures. The authors have made an interesting contribution in the area of risk estimating by presenting a simple scoring method for analysis of the perceived probability of risk occurrence and degree of risk impact. Although the study is a good solution to the problem at hand, the method presented is not without problems. The dis￾cusser feels several procedural questions arise from the paper. Let us begin with a brief examination of how the risk signifi- cance index is estimated in the paper. The index denoted as RS contains two attributes: ~1! the probability level of the risk occur￾rence ~a! and ~2! the degree of impact or the level of loss if the risk occurs ~b!. Thus, in the author’s paper, for each respondent i the risk significance is given by Sj i 5aj i bj i As indicated by the authors, the model for the calculation of index score for risk k is, RS~k!5( j51 54 Sj i /54 (1) In this case, with an impact value ~b! and a probability of occur￾rence ~a!, the expected value of the impact is actually E(ab) 5aE(b). However, the problem is that both a and b are ‘‘per￾ceived’’ values. Using this approach, the expected value of the product ab should be E~ab!5E~a!E~b!1Cov~a,b! (2) where Cov~a,b! represents the covariance between a and b such that Cov(a,b)5S@(a2¯a)(b2¯ b)/54#, and ¯a and ¯ b represent the mean values of a and b respectively. In addition, we have to assume that a and b are independent, that is, how the respondents perceive the probability that risk occurrence is not affected by its corresponding perceived degree of risk impact. This means that Cov(a,b)50, implying E(ab) 5E(a)E(b). Following this way, the correct calculation of the index score should be RS~k!5S ( j51 54 aj i /54D S ( j51 54 bj i /54D (3) Thus, the risk significance index would be underestimated using Eq. ~1!. The authors should first separately calculate the average values of a and b for each risk item. The other problem is that a and b may not be independent. However, the problem would be encountered if a high degree of loss tends to be associ￾ated with a low probability of occurrence. In this case, we assume RS5a(b)b. Therefore, dRS/db5a1b(da/db). This suggests that the change of RS in response to a change in b can be positive or negative, depending on the value of da/db. If Cov(a,b) Þ0, it is not difficult to estimate Cov~a, b! using the existing data. With a sufficiently large sample, the author may examine Cov~a, b! for each risk item. One would expect that Cov~a, b! is negative. Because the variables take a discrete value, such that high 51.0, average50.5, and low50.1, the domain of the risk signifi- cance index should lie in the range @0.01, 1#. This means that the significance index that has a value exceeding 0.25 can be classi- fied as ‘‘high risk.’’ Alternatively ‘‘low risk’’ is indicated if the index is less than 0.25. In this case, the t test can be applied to the null hypothesis RS50.25. Based on the t test, the authors are advised to categorize the risk significance index into three levels: low risk for RS,0.25, average risk for RS not significantly dif￾ferent from 0.25, and high risk for RS.0.25. Alternatively, it is more appropriate to use frequency analysis for such study. The authors should indicate the respective fre￾quency ~in percentage! of the high, average, and low risk of each risk item. The frequency data shown in Table 1 shows that the score for each risk item is calculated as Sf a f b ~impact factor!. In Table 1, the score50.330.231.010.530.630.510.230.230.1 50.214. Table 1. Frequency Analysis Frequency a ~%! b ~%! Expected value High 30 20 0.06 Average 50 60 0.15 Low 20 20 0.004 Score — — 0.214 DISCUSSIONS AND CLOSURES 194 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2002