Fondecyt project
INFERENCE IN INSTRUMENTAL VARIABLES MODELS WITH HETEROSKEDASTICITY
Project number 11140433
Embedded Files
SUMMARY OF THE PROJECT
Instrumental variable techniques are widely used in statistics, econometrics and epidemiology in particular to address the problems associated to the omitted variable bias, simultaneity and measurement errors. Nonetheless, it is well known that certain instrumental variable methods suffer from very poor finite sample properties. In the last thirty years a growing body of literature has recognized the crucial role of the quantity and the quality of such instruments. As a consequence, a number of alternative methods have been introduced including many (weak) instrument consistent estimators, modifications of the Anderson-Rubin statistic and of the Sargan statistic just to mention a few. The result of introducing heteroskedasticity in the picture is not obvious as it is documented that in the context of estimation it may cause even many instrument robust estimators to be inconsistent. Since heteroskedasticity is a relevant situation in practice, for example in the context of cross-sectional data, we find useful to investigate a set of test statistics that may be robust in the presence of many potentially weak instruments and heteroskedasticity. We expect to come up with a series of test statistics that generalize to the case of heteroskedasticity known tests and that enjoy better finite sample properties than the existent heteroskedasticity robust tests. We claim that the tests proposed in this project would work in the case of many (weak) instruments and heteroskedasticity and would improve the finite sample properties of existing tests. The objective of the project is then to elaborate a set of test statistics that generalize some well known tests such as the Sargan overidentification test and the Anderson-Rubin (AR) test. We aim at studying their asymptotic properties and their behaviour in finite samples. Since there are not very many software that implement robust instrumental variables methods we will construct a package for the free software R that will include our findings and other classical and more modern instrumental variables estimators and tests.
BUDGET: 60.303.000 Chilean Pesos
DURATION OF THE PROJECT: 2014-2016
OUTPUT, WORK IN PROGRESS AND RELATED PUBLICATIONS
Bilinear Form Test Statistics for Extremum Estimation (with F. Osorio), 2020, Economics Letters.
Inference in Instrumental Variables Models with Heteroskedasticity and Many Instruments (with G. Mellace and Z. Sándor), 2020, Econometric Theory.
Innovation and Cost Efficiency in the Banking Industry: the Role of Electronic Payments (with G. Ardizzi and C. Petraglia), 2019, Economic Notes.
On the Finite Sample Properties of Conditional Empirical Likelihood Estimators (with Z. Sándor), 2017, Communications in Statistics - Simulation and Computation, 46, 1520-1545.
THESIS STUDENTS
Freddy Lopez (PhD Statistics, PUCV, UV, UTFSM) Contributions to the sensitivity analysis for instrumental variables regression and other topics.
Luis Acosta Romero (MSc Statistics, PUCV) Rendimiento en Establecimientos de Educacion Basica Utilizando Variables Instrumentales
Marcelo Montoya Mancilla (BSc Statistics, PUCV) Variables Instrumentales Aplicadas a la Salud
CONFERENCE
Session on "Instrumental Variables: Theory and Applications" at the 11th International Conference of the ERCIM WG on Computational and Methodological Statistics in Pisa (Italy) between the 14th and 16th of December 2018.
Instrumental variable techniques are widely used in statistics, econometrics and epidemiology in particular to address the problems associated to the omitted variable bias, simultaneity and measurement errors. Nonetheless, it is well known that certain instrumental variable methods suffer from very poor finite sample properties. In the last thirty years a growing body of literature has recognized the crucial role of the quantity and the quality of such instruments. As a consequence, a number of alternative methods have been introduced including many (weak) instrument consistent estimators, modifications of the Anderson-Rubin statistic and of the Sargan statistic just to mention a few. The result of introducing heteroskedasticity in the picture is not obvious as it is documented that in the context of estimation it may cause even many instrument robust estimators to be inconsistent. Since heteroskedasticity is a relevant situation in practice, for example in the context of cross-sectional data, we find useful to investigate a set of test statistics that may be robust in the presence of many potentially weak instruments and heteroskedasticity. We expect to come up with a series of test statistics that generalize to the case of heteroskedasticity known tests and that enjoy better finite sample properties than the existent heteroskedasticity robust tests. We claim that the tests proposed in this project would work in the case of many (weak) instruments and heteroskedasticity and would improve the finite sample properties of existing tests. The objective of the project is then to elaborate a set of test statistics that generalize some well known tests such as the Sargan overidentification test and the Anderson-Rubin (AR) test. We aim at studying their asymptotic properties and their behaviour in finite samples. Since there are not very many software that implement robust instrumental variables methods we will construct a package for the free software R that will include our findings and other classical and more modern instrumental variables estimators and tests.
BUDGET: 60.303.000 Chilean Pesos
DURATION OF THE PROJECT: 2014-2016
OUTPUT, WORK IN PROGRESS AND RELATED PUBLICATIONS
Bilinear Form Test Statistics for Extremum Estimation (with F. Osorio), 2020, Economics Letters.
Inference in Instrumental Variables Models with Heteroskedasticity and Many Instruments (with G. Mellace and Z. Sándor), 2020, Econometric Theory.
Innovation and Cost Efficiency in the Banking Industry: the Role of Electronic Payments (with G. Ardizzi and C. Petraglia), 2019, Economic Notes.
On the Finite Sample Properties of Conditional Empirical Likelihood Estimators (with Z. Sándor), 2017, Communications in Statistics - Simulation and Computation, 46, 1520-1545.
THESIS STUDENTS
Freddy Lopez (PhD Statistics, PUCV, UV, UTFSM) Contributions to the sensitivity analysis for instrumental variables regression and other topics.
Luis Acosta Romero (MSc Statistics, PUCV) Rendimiento en Establecimientos de Educacion Basica Utilizando Variables Instrumentales
Marcelo Montoya Mancilla (BSc Statistics, PUCV) Variables Instrumentales Aplicadas a la Salud
CONFERENCE
Session on "Instrumental Variables: Theory and Applications" at the 11th International Conference of the ERCIM WG on Computational and Methodological Statistics in Pisa (Italy) between the 14th and 16th of December 2018.
Google Sites
Report abuse