Cross-file working

• Stats programs usually work with rectangular files: $N$ rows $\times$ $M$ variables
• Individual files in the BHPS database are rectangular and can be directly analysed
• However, more interesting analyses need to work across these files

• For instance, to track attitudes over time, or to look at income change
• This is the basic panel operation

• For instance, to calculate the number of employed persons in the household

• This is a very common operation
  • apply household data to individual
  • apply individual data to multiple job spell file
  • apply aggregated data back to lower level

• We may wish to generate a file of couples - which will be symmetric
• or parents with children - which will not