There are some very exciting things and some very worrying things about this proposal. First, to start with the exciting aspects, the use of panel data by students and researchers in their projects is already taking off at rapid speed now that a number of longitudinal panel surveys are available online. I have met countless academics who have told me recently that they commonly direct their students to panel studies such as the BHPS instead of asking them to design and collect survey data of their own. So we should be confident of their being a good market for a practical guide to using panel data in research. Secondly, Essex University is the THE centre for panel survey expertise in Europe and so the fact that these two authors are part of such a highly-respected team will help sales of the book. But, as flagged above, there is a big ‘BUT’ to this proposal, and that is the authors' dogged determination to support this book with Stata software, rather than with, ideally, SPSS or R. Stata is not widely used in the UK in the social sciences and I fear there could be an impact on sales if stata is too prominent. The authors have agreed to include an appendix on R and will have some coverage of how to use SPSS in analysing panel data on a modest website. Other changes to the chapter structure make the book more accessible and practical, and the agreement to include a range of international panel studies in the guide will help overseas sales. But there is no getting around the fact that the stata dimension is far from helpful. As a consequence I have sought to control costs and I would suggest Indian printing and a very modest royalty offer. It is also I think a Mod in it's market potential, but will have a sales pattern more characteristic of a supp. I want a Guide to Panel Data to support the list, and Essex is the ideal department to supply authors for such a text, but this is not quite the ideal book.

Analysis of Panel Data for Continuous Dependent Variables

Aim

This chapter discusses estimation methods for panel data where the dependent variable is continuous. It shows how to estimate pooled models and how to account for individual unobserved heterogeneity; how to compute different types of standard errors, linear and non-linear tests for model selection; and how to obtain predictions, residuals, etc. The techniques are illustrated using as examples wage equations estimated using BHPS data for the period 1991-2008.

Introduction

The most important characteristic of panel data is that we have repeated observations per individual. Following individuals over time allows us to better disentangle causality and to better control for unobserved individual heterogeneity (that is, differences across individuals that do not change over time and which are not ...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles