Participation in the 24th Behavioral Modeling Summer School
From September 22 to 24, the 24th Behavioral Modeling Summer School was held in a hybrid format, both in person in Univ. Tokyo and online. Six students from our laboratory—one first-year master’s student and five fourth-year undergraduates—took part in the event.
This year, participants were provided with a newly developed panel dataset that traces residential transitions over more than a decade in the three Tohoku prefectures (Iwate, Miyagi, and Fukushima), including data from individuals affected by the Earthquake.
Our team took on the challenge of analyzing this dataset by constructing a residential location choice model using a multinomial logit (MNL) framework, with the goal of exploring policy measures to mitigate population outflow to Tokyo.
While we were able to identify variables characterizing migration to Tokyo relatively easily, pinpointing factors associated with “staying in Tohoku”—such as regional attachment—proved to be extremely challenging, making the analysis quite demanding.
Nevertheless, through repeated trial and error and close teamwork within the limited time available, we continuously refined our model and carefully interpreted its parameters.
As a result, our team was awarded sixth place overall among all participating groups.
By boldly tackling the challenging task of analyzing difficult data, this experience proved invaluable. It taught us both the depth of behavioral models and the essential lesson that persistent effort is indispensable for building good models.
We would like to express our sincere gratitude for providing us with such a wonderful opportunity.
[Ito]
