The ex-ante evaluation of policies using structural microeconometric models is based on estimated parameters as a stand-in for the truth. This practice ignores uncertainty in the counterfactual policy predictions of the model. We develop an approach that deals with parametric uncertainty and properly frames model-informed policy-making as a decision problem under uncertainty. We use the seminal human capital investment model by Keane and Wolpin (1997) as a well-known, influential, and empirically-grounded test case. We document considerable uncertainty in their policy predictions and highlight the resulting policy recommendations from using different formal rules on decision-making under uncertainty.
Collaborators: Janos Gabler, Lena Janys
Economists often estimate a subset of their model parameters outside the model and let the decision-makers inside the model treat these point estimates as-if they are correct. This practice ignores model ambiguity and opens the door for model misspecification and post-decision disappointment. We develop a framework to explore and evaluate decision rules that explicitly account for the uncertainty in the first step estimation and assess their performance in a decision-theoretic setting. We show how to operationalize our analysis by studying a stochastic dynamic investment model where the decision-maker takes ambiguity about the model’s transition dynamics directly into account.
Collaborators: Maximilian Blesch
Standard consumption utility is linked in time to a consumption event, whereas the timing of prosocial utility flows is ambiguous. Prosocial utility may depend on the actual utility consequences for others – it is consequence-dated – or it may be related to the act of giving and is thus choice-dated. Even though most prosocial decisions involve intertemporal trade-offs, existing models of other-regarding preferences abstract from the time signature of utility flows, limiting their explanatory scope. Building on a canonical intertemporal choice framework, we characterize the behavioral implications of the time structure of prosocial utility. We conduct a high-stakes donation experiment that allows us to identify non-parametrically and calibrate structurally the different motives from their unique time profiles. We find that the universe of our choice data can only be explained by a combination of choice- and consequence-dated prosocial utility. Both motives are pervasive and negatively correlated at the individual level.
Collaborators: Felix Chopra, Armin Falk, Thomas Graeber
Uncertainty quantification and robust decision-making: Initiating a transdisciplinary research program
Computational models play an ever-increasing role in informing decisions. Domain-expertise is essential for developing models tailored to their intended application and the available data. However, the shared need to calibrate models to data and enable model-informed decisions creates many transdisciplinary research opportunities. Uncertainty, for example, is a major challenge across scientific domains. However, in practice, we often display incredible certitude when analyzing our models’ implications and disregard the uncertainties involved. Consequently, we accept fragile findings as facts, dueling certitudes stifle constructive debate, and we do not identify gaps in our knowledge. We bring together a transdisciplinary research team from economics, epidemiology, and finance to address these shortcomings.
We present background material on a class of structural microeconometric models to facilitate transdisciplinary collaboration in their future development. We describe the economic framework, mathematical formulation, and calibration procedures for the so-called Eckstein-Keane-Wolpin (EKW) models. We provide an exemplifying analysis of the seminal model outlined in Keane & Wolpin (1997) and present our group’s ensemble of research codes that allow for its specification, simulation, and calibration. We summarize our efforts drawing on research outside economics to address the computational challenges in applying EKW models and improve their results’ reliability and interpretability.