Wonderful news! Congratulations!

Thanks to @econometric.bsky.social for the invitation!

#EconomicsofEducation #Expectations #StructuralModels #StructuralEconometrics #SubjectiveExpectations

💡Belief data = a powerful tool in structural models of education choices.
They help us understand how people think—and how belief errors shape behavior and education policy outcomes.

Enrico Miglino & I use belief data on effort returns to estimate a dynamic model of pre-college effort.
Belief errors help explain RCT results on preferential admissions.
Simulations suggest: info interventions could reduce disincentive effects of preferential admissions.

4️⃣ To quantify belief errors + welfare impacts
Kapor, Neilson & Zimmerman (2020):
Belief errors about admissions can reverse welfare comparisons across school assignment mechanisms.

Wiswall & Zafar find smaller earnings of major choice elasticities than past work. Why?
Because people who think a major pays more also tend to like it more.
Beliefs and preferences correlate.
Info experiments + belief data help us purge this bias.

3️⃣ To study learning
Beliefs change. Panel data lets us see how they change—and the effects on behavior.
📚 Stinebrickner&Stinebrickner (2014): beliefs about STEM aptitude evolve slowly.
📚 Wiswall & Zafar (2015): info experiments reveal causal effects of beliefs.

2️⃣ To validate the model
Use beliefs on stated choice probabilities to test model fit.
Do model-predicted choices ≈ survey-elicited choices?
🌟 Gonul & Wolpin (1985);
🌟 Delavande & Zafar (2019): validate university choice model using Pakistani student data.

1️⃣ To inform the model
Arcidiacono, Hotz, Maurel & Romano (2020):
Students' beliefs about earnings + occupational choices help test:
Do they act on expected returns?
(ATT > ATE?)
Answer: partly. Non-monetary factors matter too.
➡️ Roy model with non-pecuniary preferences.

We’ve made major progress in understanding how people form beliefs—and how we can use those beliefs in our models.

In the lecture, I outlined 4 ways belief data can be used in structural econometrics:

Education decisions are dynamic.
➡️ Go to college = lose earnings today, hope for better returns later.
But those returns are uncertain.
Structural models assume people form expectations and maximize expected utility.
But how accurate are those expectations?