Accessing the New Earnings Survey Panel Dataset : efficient techniques and applications
The New Earnings Survey Panel Dataset is one of the largest datasets of its kind in the world. Its size and confidentiality restrictions present considerable difficulties for analysis using standard econometric packages. This thesis presents a number of methods for accessing the information held within the panel relatively efficiently, based upon the use of cross-product matrices and on data compression techniques. These methods allow, for the first time, the panel aspect of the dataset to be used in analysis. The techniques described here are then employed to produce an overview of changes in the UK labour market from 1975 to 1990 and detailed estimates of male and female earnings over a fourteen year period. These are the first panel estimates on the dataset, and they indicate the importance of allowing the parameters of any labour market model to vary over time. This is significant as panel estimators typically impose structural stability on the coefficients. A comparison of cross-section and panel estimates of earnings functions for males indicate that the allowance for individual heterogeneity also has a notable effect on the estimates produced, implying simple cross-sections may be significantly biased. Some preliminary estimates of the male-female wage gap indicate that variation over time has an important part to play in accounting for the differences in wages, and that "snapshot" studies may not capture dynamic changes in the labour market. Individual differences also playa significant role in the explanation of the wage gap.