Simulated Maximum Likelihood Estimation of Weitzman's Sequential Search Model

This webpage shares the Matlab script that estimates the sequential search model by Weitzman (1979).

Settings:

• Consumers search through products to resolve the uncertainty in the normally distributed match values. 

• Consumers make the search and purchase decision by Weitzman (1979)'s sequential search algorithm. 

• For each search, consumers incur a search cost that is assumed to follow an exponential distribution. 

• Search cost is assumed to vary at the consumer-product level and to be unobserved by researchers.

What the script does:

"main_for_loop.m" file and "main_vectorize" file simulate the product characteristics and the search and purchase decisions by 1,000 consumers and estimate the sequential search model by Weitzman. 

Instruction:

Download and extract the zip file below. Run "main_for_loop.m" file or "main_vectorize" file.

"main_for_loop.m" utilizes for-loop to construct the simulated likelihood. 

"main_vectorize" executes pre-computation to vectorize the likelihood function evaluation and remove the for-loop. 

Note:

"search_function" folder needs to be placed in the same directory as "main_for_loop.m" script and "main_vectorize" file.

References:

Weitzman, Martin L. “Optimal Search for the Best Alternative.” Econometrica, vol. 47, no. 3, 1979, pp. 641–54. 

Kim, Jun B., et al. “Online Demand Under Limited Consumer Search.” Marketing Science, vol. 29, no. 6, 2010, pp. 1001–23. 

McFadden, Daniel. “A Method of Simulated Moments for Estimation of Discrete Response Models Without Numerical Integration.” Econometrica, vol. 57, no. 5, 1989, pp. 995–1026. 

Zhenling Jiang, Tat Chan, Hai Che, and Youwei Wang. "Consumer Search and Purchase: An Empirical Investigation of Retargeting Based on Consumer Online Behaviors." Marketing Science, 2021 40:2, 219-240 

Yuxin Chen, Song Yao (2016) Sequential Search with Refinement: Model and Application with Click-Stream Data. Management Science 63(12):4345-4365. 

Updates:

2023.03.12: Uploaded a Matlab script that uses the for loop in the estimation with the new method, KSFS, CFS, and Jiang et al. (2021)'s  method

2023.09.16: Uploaded a Matlab script that uses the vectorization pre-computation with the proposed method, KSFS, CFS, and Jiang et al. (2021)'s  method

Download Code

Matlab Code