João Cruz

Efficient Estimation and Inference with Binary Response Dyadic Regression

Dyadic data is known to exhibit a distinctive form of correlation known as dyadic clustering. To study binary network data, several latent variable models have been developed to account dependence, yet the latent structures proposed can be rather complex and valid inference depends on its correct specification.

On the other hand, the use of fixed effect procedures although robust to potential endogeneity issues, controls out relevant information in the model and impedes the identification of coefficients associated with agent-specific regressors. We propose to work with minimal assumptions on the data generating process, focusing on the first two data moments. This allows to construct a consistent and asymptotically normal estimator based on a quasi-likelihood procedure accounting for dyadic clustering. The proposed estimator has the advantage of not depending on strong assumptions on the specification of the latent components, while aiming to gain efficiency in the estimation process and inference.

Show more

The Gender Pay Gap in STEM – An Empirical Analysis for Portugal

At a submission process to the International Labour Review

Sectors such as science, technology, engineering and math (STEM) present a structural gender pay gap (GPG). In the present research we use official data from Portuguese earnings to investigate the GPG between 2010 and 2021 in these sectors.

We use decomposition methods to estimate the explained and unexplained components of GPG at the mean and at the distribution quartiles. Results obtained show that GPG behaves heterogeneously across STEM fields. In occupations related to Information and communication technologies (ICT) included in STEM, GPG is becoming more severe in recent years contrasting with the pattern for the overall labour force.

Show more