Publications
Pérez Izquierdo, T. and Pronkina, E.
Economics of Transition and Institutional Change
2023
Behind the curtain: How did women's work history vary across Central and Eastern Europe?
This paper investigates the differences in female work experience across Central and Eastern European countries (CEECs). We use retrospective SHARELIFE data to analyse women's work history from 1950 to 1990. We provide descriptive evidence that women's work experience varied across CEECs. Furthermore, we argue that comparing the former provinces of the Russian Empire in Lithuania and Poland provides a natural experiment, allowing us to disentangle the effect of the differential implementation of the Soviet regime from the pre-existing differences. We find that during communism, Lithuanian women worked 2 years more by age 50 relative to their Polish counterparts. This effect is one-third of that found in the East-West Germany comparison. We propose several potential mechanisms behind this finding: the degree of land collectivization, the Church's influence and the sectoral composition. Accordingly, this study's findings highlight the importance of country differences in CEECs.
Working papers
Cuerno, M.; Galáz-García, F.; Galáz-García, S.; and Pérez Izquierdo, T.
Reassessing Relationality for Bipolar Data
Methods for clustering people into construals--social affinity groups of individuals who share similarities in how they organize their outlooks on a collection of issues--have recently gained traction. Relational Class Analysis (RCA) is currently the most commonly used method for construal clustering. RCA has been applied to identify affinity groups in social spheres as varied as politics, musical preferences, and attitudes towards science. In this study, we highlight limitations in RCA's ability to accurately identify the number and underlying structure of construals. These limitations stem from RCA's mathematical underpinnings and its insensitivity to the bipolar structure of the survey items, which require respondents to place themselves in a support or rejection space and then express the intensity of their support or rejection. We develop an alternative method, which we call Bipolar Class Analysis (BCA), that aims to address this foundational limitation. BCA conceptualizes people's attitudinal positions as moving along support/rejection semispaces and assesses similarity in opinion organization by taking into account position switches across these semispaces. We conduct extensive simulation analyses, with data organized around different construals, to demonstrate that BCA clusters individuals more accurately than RCA and other available alternatives. We also replicate previous analyses to show that BCA leads to substantively different empirical results than those produced by RCA in its original and later versions, and by Correlational Clustering Analysis (CCA), a method that has been proposed as an alternative to RCA.
Escanciano, J. C. and Pérez Izquierdo, T.
Automatic Locally Robust Estimation with Generated Regressors
Many economic and causal parameters of interest depend on generated regressors. Examples include structural parameters in models with endogenous variables estimated by control functions and in models with sample selection, treatment effect estimation with propensity score matching, and marginal treatment effects. Inference with generated regressors is complicated by the very complex expression for influence functions and asymptotic variances. To address this problem, we propose Automatic Locally Robust/debiased GMM estimators in a general setting with generated regressors. Importantly, we allow for the generated regressors to be generated from machine learners, such as Random Forest, Neural Nets, Boosting, and many others. We use our results to construct novel Doubly Robust and Locally Robust estimators for the Counterfactual Average Structural Function and Average Partial Effects in models with endogeneity and sample selection, respectively. We provide sufficient conditions for the asymptotic normality of our debiased GMM estimators and investigate their finite sample performance through Monte Carlo simulations.
Pérez Izquierdo, T.; Barrio, I.; and Esteban, C.
Dynamic prediction of death risk given a renewal hospitalization process
Predicting the risk of death for chronic patients is highly valuable for informed medical decision-making. This paper proposes a general framework for dynamic prediction of the risk of death of a patient given her hospitalization history, which is generally available to physicians. Predictions are based on a joint model for the death and hospitalization processes, thereby avoiding the potential bias arising from selection of survivors. The framework accommodates various submodels for the hospitalization process. In particular, we study prediction of the risk of death in a renewal model for hospitalizations, a common approach to recurrent event modelling. In the renewal model, the distribution of hospitalizations throughout the follow-up period impacts the risk of death. This result differs from prediction in the Poisson model, previously studied, where only the number of hospitalizations matters. We apply our methodology to a prospective, observational cohort study of 401 patients treated for COPD in one of six outpatient respiratory clinics run by the Respiratory Service of Galdakao University Hospital, with a median follow-up of 4.16 years. We find that more concentrated hospitalizations increase the risk of death.
Pérez Izquierdo, T.
The determinants of counterfactual identification in the binary choice model with endogenous regressors
The Counterfactual Average Structural Function (CASF) is the Average Structural Function (ASF) averaged with respect to a counterfactual distribution of covariates. In the binary choice model, the CASF measures the rate of successes in a counterfactual scenario. This paper shows that the CASF is non-parametrically identified as a weighted average, where the weights are given by a likelihood ratio. Estimation of the CASF at the regular root n-rate requires a square integrability condition for those weights. This necessary condition depends on instrument strength, the degree of endogeneity, and the relevance of the regressors. Much insight is gained from the single normal regressor model. In this setting, I show that extrapolation strength (the capacity of the model to regularly identify the CASF) increases with the relevance of the regressor and, surprisingly, with the degree of endogeneity. The impact of instrument strength is found to be non-monotone. Moreover, I find a threshold for instrument strength below which regular identification of the CASF is not possible.