Flexibility of Transport Choice in a Real-Option Setting: An Experimental Case Study

Abstract

The authors conducted an experimental study about road haulers' choice between railways and roads. If road haulers choose to load their truck on a train, they are not able to switch to using the road again until they reach the end of the line. On the contrary, highway choice is flexible in that the road hauler can change road at any time. Building upon the irreversibility of one of the two choices, the authors build a model in which traffic levels are risky using both infrastructures. In the model, they found that haulers' choices depend on the infrastructure price for railways and on the information level obtained by the hauler during travel. The flexible option is more valuable the more informed the agent is about traffic levels. These theoretical predictions are tested by implementing two experimental treatments. In the first, agents gain no information during their travel; in the second they become perfectly informed about traffic conditions during the travel. It is important that experimental payoffs are calibrated upon transport data about time and operating costs and infrastructure tariffs in France. From the experiment, the authors found that more risk-averse agents assign more value to the flexible option. The second result is that participants overreact to infrastructure price changes. The results indicate that, as the model stands, price levels and information level are important explanatory variables of choice. The higher the price is for railways and the higher the information is, the more participants give value to flexible option (highway), especially if participants are risk averse. The second result is that participants overreact to infrastructure price modification. The authors observed that, except for low levels of price, participants tend to choose the flexible option more frequently, even if it is suboptimal.

Keywords:

• preference for flexibility
• quasi-option value
• experimental economics
• travel behavior
• traffic information

Notes

¹For example, if a participant has to choose between a B lottery, say 50% to win $1 and 50% to win $90, and a certain reward R of $20, if she chooses B, random choice of a number will be made by a computer, giving $1 if the number is between 1 and 50, and $90 if the number is between 51 and 100.

²However, it is fair to say that experimental economics could induce more rational behavior than what would be observed in real life.

³As noted by Dixit and Pyndick (1995), this option value had been introduced by Arrow and Fisher (1974) and Henry (1974). This concept gives important developments over the recent years to environmental economics, following Kolstad (1992) and Hendricks (1992). For a discussion of this concept, see Fischer (2000).

⁴See Ministère des Transports (1998). In France, the official recommendation is to consider 30€ per hour. However, for other studies (e.g., Commissariat Général du Plan, 2001), this cost should vary from 4.5 to 38€/h per hour, which means an average of 21€ per hour. For our experiment, we choose 120 points per hour (i.e. 28.4€). The study of the Institut National de la Statistique et des Études Économiques (2004) estimated average price per kilometer for a truck around 1,259€/km with an average load rate of 75%. That gives revenue around 378€ for a route of 300 km (which corresponds to 1,600 points in our experimental calibration). Then, we choose 120 points; that is, 28.4€ per hour as time cost.

⁵Indeed, the generalized transport cost = (time value) × (travel time) + (average private cost) + fee. In the design, the average generalized transport cost is 1,600 points, and then the average payoff should be 1,600 points, which implies that players can have losses or gains.

⁶For that, we use the methodology of Levy and Levy (2001), on the basis of first- and second-order stochastic dominance criteria. An appendix is available from http://sites.google.com/site/laurentdenantboemont

***Significant at 10%.

**Significant at 5%.

***Significant at 1%.

***Significant at 1%.

⁷The variable context, which equals to “0” or “1,” refers to the kind of instructions we used. In one case, we used transport-framed instructions (context = “1”), and in the other case lottery-based instructions (context = “0”).

***Significant at 10%.

**Significant at 5%.

⁸In our example, she (Participant 1) chose to have an expected gain of 10 instead of maximum expected gain of 60; that is, the expected opportunity cost of flexibility is 50.

⁹However, such a conclusion should be taken with care, because revealed WTP is in most cases under theoretical WTP is the consequence of the calculus method, on the basis of binary choice. It implies that revealed WTP are restricted by expected theoretical payoffs of the two choices; that is, revealed WTP depends on theoretical expected utilities, which constraint the possible values to a narrow range.

***Significant at 1%.

***Significant at 1%.