Discussing
Social Networks and Geographic Mobility by Milena Almagro, Olivia Bordeu and Gregorio Caetano


Florian Oswald

Urban Economics Association Meeting Bocconi 2023

5 May, 2023

Overview and Findings of the Paper

Overview

  1. Social Networks (SN) matter for migration. SN \(\approx\) Amenity.
  1. SNs are important for decision to stay/leave and also for where exactly to go.

This Paper

  1. Focuses on the stay/leave part, and a particular form of SN: relatives who assist with childcare.
  1. Documents who relies more on this SN, and how migration decisions are impacted by its existence and extent.

Findings

  • Low-income HH are less mobile than high-income HH.
  • Relatives provide 70% of childcare for low-income HH (50% for high-income).
  • Presence of children in HH reduces cross-state mobility significantly.
  • Income strongly predicts amount of childcare provided by market vs relatives.
  • Correlation between mobility and reliance on family-provided childcare.

Comments

Relevant Additional Literature


  1. Gizem Kosar, Tyler Ransom and Wilbert van der Klaauw - Understanding migration aversion using elicited counterfactual choice probabilities : Direct elicitation of migration probabilities in relation to social network.


  1. Joshua Blumenstock, Guanghua Chi and Xu Tan - Migration and the Value of Social Networks: Role of social network, information vs economic support.

Everybody is equally fertile

BUT

  • I encourage you to precisely spell out the consequences of assuming exogenous fertility, and the precise assumptions you are making. In the data, people choose location and fertility jointly.

What is a Location?

  • a tuple (wage, price of childcare type j, relative price of C)
  • all generated outside of the model - there is no mechanism in the model to set prices and wages.
  • In the real world some places have very limited supply of market childcare - rural areas. Why not incorporate those constraints explicitly?
  • Locations differ starkly by housing costs for people with and without children - absent from the model. (My bias from above.)

Stylized Fact and Table 1

Table 1 this paper

Table 4 Oswald 2019. (Probit model.)

Calibration/Estimation

  • Calibration of opportunity cost to (observed) mothers’ wages: multi-way Selection problem?!
    1. Using observed wage for non-participants yields biased estimates (Heckman)
    2. high wage mothers move to expensive city (Urban literature)
    3. might have lower fertility. (?)
  • Effect of tenure: cost of childcare after 7 years of tenure might drop because…the kid is now 7+ years old!

Counterfactuals and the Lucas Critique?

  • The plan is to use the estimated model to perform different policies.
  • The model is in partial equilibrium: all prices are estimated from data, there is no mechanism to set them inside the model.
  • Policy will change allocation of people across space, presumably changing prices, how will we interpret the experiments?

Suggestions

Suggestions

  1. Focus more on the childcare story (starting in title?)
  2. The choice of different childcare mixes adds a lot of complexity - try to make more out of this! Argue that we need this level of complexity (figure 6).
  3. Maybe link somehow to gender-gap-location literature? You are providing a childcare choice model for work like Le Barbanchon, Rathelot and Roulet - Gender Differences in Job Search: Trading off Commute against Wage

End

Thanks for nice paper!

UEA 2023 Florian Oswald, SciencesPo Paris