Analysis of "Foreign STEM Workers and Native Wages and Employment in U.S. Cities"

Summary

In May of 2014, an NBER Working Paper titled "Foreign STEM Workers and Native Wages and Employment in U.S. Cities", written by Giovanni Peri, Kevin Y. Shih, and Chad Sparber, was released. Then on May 22, 2014, the Wall Street Journal posted an online article titled "Skilled Foreign Workers a Boon to Pay, Study Finds". The article states the major findings of the study as follows:

They found that a one-percentage-point increase in the share of workers in STEM fields raised wages for college-educated natives by seven to eight percentage points and wages of the noncollege-educated natives by three to four percentage points.

To illustrate the first of these findings, the article included a chart. One of the study's authors, Giovanni Peri, provided me with the spreadsheet on which the chart is based.

This analysis finds significant methodological problems with the chart and the finding that it is meant to support. The analysis also finds some evidence countering the study's finding that extending visas to more STEM workers does not affect the employment of other groups.

Key Issues with the Wall Street Journal Article and Chart

1. The chart and spreadsheet are missing San Jose, CA, the heart of Silicon Valley, which had the largest influx of foreign STEM workers from 1990 to 2010.
2. The chart and spreadsheet are missing Stamford, CT, which had the largest increase in real native college-graduate wages from 1990 to 2010.
3. The omission of San Jose, CA and Stamford, CT are not mentioned in the article. In fact, the article states that the "areas with the biggest influx of foreign STEM workers were Austin, Texas; Raleigh-Durham, N.C.; Huntsville, Ala.; and Seattle".
4. The article states that the study "examined wage data and immigration in 219 metropolitan areas from 1990 to 2010". However, the spreadsheet and chart contain data for 280 metropolitan areas. See here for more information.
5. Of the 280 metropolitan areas in the chart, 44 do not contain data for 1990 and are colored blue in the plots that replicate the WSJ chart. Their data appears to have come from a source other than the IPUMS source mentioned in the study.
6. Other metropolitan areas, such as Austin, had very different values than was calculated using IPUMS. It may be that this data was likewise calculated using data from an undefined source.
7. Mr. Peri provide me with the spreadsheet for the chart but did not provide the source of the chart's data.

Key Issues with the Study

1. I requested but was unable to obtains the data and calculations required to exactly replicate the study.
2. The study is a working paper and does not appear to have been peer-reviewed or independently replicated and/or verified.
3. Four other issues are listed in the summary of an initial attempt to replicate the study at http://econdataus.com/jole_pss.htm#section13.

Analysis

1. Findings Reported by Wall Street Journal Article
2. Chart Included to Illustrate First Finding
3. Replicating the Chart With IPUMS Data
4. Differences Between the Original and Replicated Chart
5. Linear Regression of IPUMS data
6. Other Considerations
7. Spreadsheet from Author Missing San Jose, CA and Stamford, CT
8. Spreadsheet from Author Includes Data from Unidentified Source
9. Conclusions

Description of R Programs

Source Code for R Programs Used in this Analysis

1. Source code for wsjchart1.R (creates Wall Street Journal replica plots from spreadsheet)
2. Source code for stem1data.R (extracts data from IPUMS files)
3. Source code for stem1plot.R (creates plots and tables from extracted data)
4. Source code for stem1labels0.R (creates labels for Wall Street Journal replica plots)
5. Source code for stem1labels1.R (creates labels for Native College Wages vs. Foreign Stem Workers plot)
6. Source code for stem1labels2.R (creates labels for Native STEM Workers vs. Foreign STEM Workers plot)
7. Csv file from spreadsheet containing values with 1 decimal place
8. Csv file from spreadsheet containing values with 3 decimal places
9. Data file containing the 219 metropolitian areas that are consistent from 1980 to 2010