Parameters Heterogeneity in a Model of Labour Supply: Exploring the Performance of Mixed Logit
Ugo Colombino, Marilena Locatelli
Abstract
Open-access reader
Ugo Colombino, Marilena Locatelli
Abstract
Open-access reader
In this note we investigate the empirical differences between the Random Utility model with fixed coefficients (Conditional Logit), and the Random Utility model with random coefficients (Mixed Logit). We consider a model of household labour supply developed for a project aimed at the evaluation of alternative Basic Income mechanisms. Data are drawn from the 1998 Bank of Italy survey of household income and wealth (SHIW 1998) and choice alternatives are generated using EUROMOD. We compare the estimates of the Conditional Logit and Mixed Logit. We also compare the respective results from simulating the effects of a Flat Tax reform. Although on average the estimates of Conditional Logit and of Mixed Logit are very close, the Mixed Logit estimates reveal that there is a significant unobserved heterogeneity of preferences. We also compare the simulations of a hypothetical Flat Tax reform. Although the differences are small, yet the results would imply different policy conclusions depending on whether Conditional Logit or Mixed Logit is adopted.
OpenAlex reports 2 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
In this note we investigate the empirical differences between the Random Utility model with fixed coefficients (Conditional Logit), and the Random Utility model with random coefficients (Mixed Logit). We consider a model of household labour supply developed for a project aimed at the evaluation of alternative Basic Income mechanisms. Data are drawn from the 1998 Bank of Italy survey of household income and wealth (SHIW 1998) and choice alternatives are generated using EUROMOD. We compare the estimates of the Conditional Logit and Mixed Logit. We also compare the respective results from simulating the effects of a Flat Tax reform. Although on average the estimates of Conditional Logit and of Mixed Logit are very close, the Mixed Logit estimates reveal that there is a significant unobserved heterogeneity of preferences. We also compare the simulations of a hypothetical Flat Tax reform. Although the differences are small, yet the results would imply different policy conclusions depending on whether Conditional Logit or Mixed Logit is adopted.
Key concepts: Mixed logit, Logit, Econometrics, Logistic regression, Economics, Discrete choice, Labour supply, Nested logit