Ideally, I would like to construct a matrix of occpations and industries (where the matrix cell contains a 1 if industry j uses occupation i and 0 otherwise). I would like to construct these matrices for as many years as possible.
Many thanks for your help!
Creating a matrix of occupation and industry can indeed be done with IPUMS data. IPUMS provides microdata, which are composed of individual records containing information collected on persons and households, as opposed to “summary” or “aggregate” data. You can use this microdata to produce estimates of national population figures.
First, you will want to create a data extract with a measure for Occupation and Industry. The coding for each of these measures is not stable over time, so there are several versions of each. I would recommend reading the User’s Note on Occupations and Industries to decide which versions to use for your analysis.
Once you have a data extract at the person-level, you will need to use a statistical package (e.g. SAS, STATA, or SPSS) to manipulate the data into (0,1) matrix form. To do this for a single year, you can collapse the data into one observation per Occupation-Industry pairing. Next, you will need to set any Occupation-Industry pairings that exist (i.e. has at least one person in the Census that reported working that Occupation within that Industry) to a value of 1. Here, you can transpose the dataset, so that Occupations run along the rows and Industries along the columns with a value of 1 in the appropriate cells. Finally, any missing values will need to be filled with 0’s. This will give you the matrix you were looking to produce. A similar process can be followed for each year of interest.
Let me know if you have any further questions about this process. I hope this helps!