I first drop observations for which information on the job variable is missing.
To arrive at our final sample we also drop observations from Scotland and Northern Ireland.
We also drop observations with missing data on the variables of interest.
From this initial sample, I drop observations with non-contributory occupations and missing values of occupation and establishment location, and drop individuals for whom educational attainment is missing.
This survey information is missing for some students and schools and we generally drop observations containing missing values for our main variables.
From this original sample I drop observations with missing or non-positive values for the number of employees, value added, or capital stock.
We drop observations from Florida, Louisiana, Mississippi, Texas, and Washington, which were affected by their own affirmative action policy changes during our sample period.
To avoid biases associated with left censoring (e.g., Heckman and Singer 1984) we drop observations where the individual's first service spell was prior to SFY 2000 (2782 observations).
In addition, because we are interested in the outcomes of participants after they have left the SIVP programme, we need to drop observations that are still recorded as being on the programme in the last month of the calendar.
We also drop observations with missing information in variables used in the analysis such as age, education, income, state of residence, state of birth, occupation, industry, and population size (see Tables 1, 2 and 3).
