Text

ETD_2126

http://hdl.rutgers.edu/1782.2/rucore10002600001.ETD.000052273

theses

I consider single- and multi-product risk-averse newsvendor models under two risk measures, coherent measures of risk and exponential utility function. Following from the typical format of a newsvendor model, I formulate the problems in the single- and multi-product cases and establish my models to take risk aversion into account. Thus, my models can capture the decision making of inventory managers at a different angle than most of literature in supply chain management. The key research questions are how the degree of risk aversion and product demand dependence structure interact with each other and affect jointly to the optimal decision of inventory managers. My models can find their applications in many manufacturing, distribution and retailing companies that handle short life-cycle products.
From my extensive literature review, I summarize and tabulate the literature of risk-averse inventory models and categorize typical approaches to risk-averse inventory models into four groups by the risk measures used. I discuss similarities and differences between the models. In particular, I provide clear axiomatic criteria to evaluate validity of risk measures in risk-averse newsvendor models. By the axiomatic criteria, coherent measures of risk are chosen to fit best for risk-averse newsvendor models, but the exponential utility function is also studied for a comparison purpose. This axiomatic approach can be also applicable to other types of risk-averse inventory models.
In the main results, I study the impact of risk aversion on the optimal ordering quantity. For single-product models, I obtain closed-form optimal ordering quantity under coherent measures of risk and closed-form approximation under exponential utility function. For a large but finite number of products, I also obtain closed-form approximations under the both risk measures when product demands are independent. My approximations are as simple to compute as the risk-neutral newsvendor solutions and the gap between the optimal solutions and approximations quickly converges to zero as the number of products increases. Then I prove that the risk-neutral solution is asymptotically optimal under coherent measures of risk, as the number of products tends to be infinity. The same proposition is proved under exponential utility function, as the ratio of the degree of risk aversion to the number of products goes to zero. Thus, in both cases, risk aversion has no impact in the limit. Demand dependence significantly affects the optimal ordering quantity. I derive analytical and numerical insights for the interplay between demand correlation and risk aversion. All these results are consistent with our insights and confirmed by numerical examples from my computational study.
I conclude my dissertation by comparing risk-averse newsvendor models and financial portfolio optimization models.

Ph.D.

Includes bibliographical references (p. 116-123)

by Sungyong Choi

doi:10.7282/T32Z15PS

ETD doctoral

The author owns the copyright to this work.

ETD

application/pdf

application/x-tar

808960

b55e8b254b4cfbf23b7536918cbd3336952ed963