1.201 / 11.545 / ESD.210 Transportation Systems Analysis: Demand and Economics Fall 2008 Midterm Exam Thursday, October 9, 2008 • You have 1.5 hours to complete this exam. • Show all your work to get full credit, and please clearly box your answers. • This is an open book and open note exam • A scientific calculator will be required to answer some of the problems on this exam Question 1 (34 points) A Freight Shipper's Mode Choice Model An independent trucking company is considering entering the lumber transport market in the Pacific Northwest, in which lumber companies currently have only two choices for how to ship their product: 1. Private Carriage (that is, ship it themselves with their own trucks) 2. Rail Freight The company has hired some consultants to help understand the current shape of the market for lumber shipping services. The first step in predicting the demand for this new trucking service is to estimate lumber companies' sensitivity to different shipping-related variables. A multinomial logit model was specified and then estimated on a dataset that included information on a number of shipping decisions made by shippers in the Pacific Northwest. Table 2.1 includes the specification of the model and the estimation results. TABLE 2.1: Model specification and estimation results ΒASC_PC ΒSIZE ΒVALUE βCOST ΒTIME ΒRELIABILITY ΒDIST_SIDING Private Carriage 1 0 0 COST_PC TIME_PC REL_PC 0 Rail Freight 0 SIZE VALUE COST_RAIL TIME_RAIL REL_RAIL DIST_ SIDING Estimated Coefficie nt 0.04 2.5 7.51 -14.14 -6.43 -3.48 -11.21Definitions of variables: ASC_PC is the alternative-specific constant for Private Carriage. SIZE is the size of the shipment in units of 10,000 pounds. VALUE is the commodity value of the shipment in dollars per pound. COST_PC is the cost of making the given shipment by Private Carriage, in units of $10,000. COST_RAIL is the cost of making the given shipment by Rail Freight, in units of $10,000. TIME_PC is the average transit time in transit of a given shipment by Private Carriage, measured in days. TIME_RAIL is the average transit time of a given shipment by Rail Freight, measured in days. REL_PC is the reliability of transit time by Private Carriage, measured in standard deviations of transit time, in days. REL_RAIL is the reliability of transit time by Rail Freight, measured in standard deviations of transit time, in days. DIST_SIDING is the distance, in miles, of the shippers' operations from a rail siding. Answer the following questions: a. What are the effects implied by the model of the following variables on an lumber company’s shipping decisions: i. Shipment Size (3 points) ii. Distance from a Siding (3 points) b. What is the value (in dollars) of one day of transit time implied by the model? What about a day of potential variability in transit time? Does the difference between these two values make sense? Explain. (8 points) c. Table 2.2 contains data on the model’s independent variables for one lumber shipment. From the data in Table 2.2 and the estimation results in Table 2.1, calculate the probability that the shipment will be made by each of the two alternatives. (9 points) Table 2.2: Data on the model’s independent variables for one example lumber shipment Variable Observed value for the individual SIZE 120,000 pounds VALUE $0.50 per pound COST_PC $4,200 COST_RAIL $3,000 TIME_PC 2.5 days TIME_RAIL 4 days REL_PC 1 day REL_RAIL 2 days DIST_SIDING 2 miles 2d. The consultants generated a sample of 1,000 lumber shipments that they believed was representative of the entire lumber shipping market in the Pacific Northwest. For each shipment in that sample, there were data on each of the two currently available alternatives (similar to the information shown in Table 2.2). How would you use these data to estimate the proportion of all shipments in the Pacific Northwest that are made by Private Carriage? (4 points) e. After the trucking service is introduced, a new model for mode choice has to be specified and estimated with three alternatives (Private Carriage, Rail Freight, Truck Freight). How would you structure such a new model? Please explain your thinking. (Hint: Think of the similarities among some of the three alternatives. Your answer can be in the form of a sketch.) (7 points) 3Question 2 (36 points) Path Choice Models and Investment on the London Underground Mike (the TA) took the London Underground to commute to work during his summer internship in London. From where he lived right near Highgate Station to where he worked downtown there were two possible ways he could go: 1. via Euston Station to Victoria Station 2. via Embankment Station to St. James Park Station Both paths involve a transfer and have similar end-to-end travel times, but he had a hard time deciding which was best. The schematic diagram of these paths with network attributes is shown in the figure below. Highgate Station (Start) Waiting Time: 3 Minutes Northern Line Travel Time: 13 Minutes Euston Station Transfer to Victoria Line: Easy Waiting Time: 1 Minute Victoria Line Northern Line Travel Time: 8 Minutes Travel Time: 10 Minutes District Line Travel Time: 3 Minutes Embankment Station Victoria Station (End 1) St. James Park Station (End 2) Transfer to District Line: Hard Walk to Office: 12 Minutes Walk to Office: 5 Minutes Waiting Time: 3 Minutes The London Underground has estimated a multinomial logit model for Path Choice across its entire network with the coefficients shown in Table 3.1: Table 3.1: Path Choice model specification and estimation results βTravel_Time βWait_Time βWalk_Time βHard_Transfers βEasy_Transfers Coefficient Estimate -0.17 -0.32 -0.37 -1.42 -0.66 (all times in this model are specified in minutes) 4This model handles transfers by labeling all transfers as either 'hard' or 'easy' depending on the complexity, walking distance, and stairs involved. Waiting time for transfers is included in the Wait_Time variable. Assume that the London Underground uses a Value of Time (VOT) of $20 per hour of in-vehicle Travel Time and that Mike and 99 other people make this same trip every day. a. Make a table of the values for both paths of the variables needed to apply the path choice logit