The immigration debate is very contentious, with “factual” claims coming from every which way. Not surprisingly, I often hear people say that “you can’t believe anything anymore because you don’t really know what the guy/gal did to reach that result.” And those suspicions are perfectly justified.
I’ve been teaching for a long time, but it wasn’t until last semester that I discovered how useful it was to show students how research gets done in real time. I first tried it out with my Mariel paper, where I could go from the raw CPS data to this striking graph showing the negative effect of the Marielitos in a few minutes with a bare minimum of statistical manipulations.
The top (blue) line gives a 3-year moving average of the weekly wage of working men outside Miami; the bottom (red) line gives the corresponding trend in Miami. I’ve now made this by-the-numbers exposition a standard part of my show whenever I present the Mariel paper at a seminar. It is far more convincing than my claiming: “This is what the data look like. Trust me!”
Some professors have told me that they would like to do something like this in their own classes. And it occurred to me that readers of this blog, many of whom have probably never seen how this type of data analysis is done, would be interested in taking a short video tour that illustrates how you can start from the raw data (publicly available at the IPUMS website); select the sample of low-skill, non-Hispanic men aged 25-59; calculate the average weekly wage of those workers in Miami and elsewhere; and, presto, end up with the graph above, documenting that something did indeed happen in Miami after 1980. Enjoy!
For the geeky reader:
1. Here’s an earlier post on why using the sample of non-Hispanic men aged 25-59 is preferable to the sample used in studies claiming that nothing happened in post-Mariel Miami.
2. The STATA commands used to derive the wage trends are in this file: Mariel Interactive Code
3. The wage trends reported in my paper use data that has been filtered further, focusing on workers in urban areas and excluding outliers with very low or very high hourly wage rates. The additional filtering doesn’t really matter all that much.
UPDATE #1: Robert VerBruggen of Real Clear Policy tweeted that it would be useful if the raw data files from the CPS were available in Excel format. Well, here they are. These files contain the relevant variables for all persons sampled between 1976 and 1991. My programming skills with Excel are extremely limited, so I have no clue about how to go from the data in these files to the graph above. Curious readers are on their own. If anyone can neatly describe the series of Excel steps that would lead to the graph, feel free to email me the information and I will add it to this post.
UPDATE #2: Robert VerBruggen sent me an email listing the steps that would allow anyone to construct the graph on their own, using a tool in the IPUMS website that I knew nothing about. This tool lets the user calculate average earnings for any particular sample of persons in any year of the CPS. The very-easy-to-follow steps are listed here; and the Excel worksheet that Robert created and that replicates the graph presented above is here. Thank you, Robert!