What do migration flows reveal about NAFTA?
Brooke Hart
Research topic
Since the implementation of the North American Free Trade Association (NAFTA) in 1994, increasing awareness of the widening trade deficit in combination with claims by the current administration have suggested that the United States has sent many jobs across the border to Mexico. In that case, the loss of jobs would undoubtedly impact the migratory patterns for regions whose industry concentration is manufacturing. This begs the questions, how have migration flows been affected since NAFTA was enacted and what does this reveal about overall employment? If jobs are in fact being lost to Mexico, it is reasonable to hypothesize that regions with the most manufacturing employment would see decreased inflow rates and increased outflow rates.
Theoretical Basis
There’s been countless studies published within the last decade attempting to assess the true impact that NAFTA has had on the economy. Critics of the deal, including the current head of administration, argue that, “it stripped us of manufacturing jobs. We lost our jobs. We lost our money. We lost our plants. It is a disaster.” (Trump - Oct. 9 in St. Louis, Mo.) According to an article published by the Federal Reserve Bank of Dallas, nearly 710,000 jobs were lost between 1994 and 2014 as a result of increased imports from Mexico and Canada or due to shifts in production. According to Brooks Jackson and FactsCheck.org, anti-NAFTA politicians advertise the loss of 1 million jobs. However, contrasting studies have concluded the trade deal resulted in much smaller job losses or even a small net gain.
NAFTA supporters highlight the creation of jobs since 1994, saying that the deal has, in reality, raised employment. In fact, the U.S. Chamber of Commerce claims that NAFTA has brought millions of jobs to the U.S.; be that as it may, those claims count only the jobs gained by increased exports. This neglects the jobs lost by increased imports.
Taking both of these arguments into account, it’s still difficult to assess the true impact that NAFTA has had. Naturally, in any given industry, there will be the elimination and creation of jobs over time, but neither of those figures are useful indications of the strengths and weaknesses of NAFTA when not looked at in context. The full story can only be told with both. Additionally, it’s hard to trust the objective nature of data when the source is often biased in one way or another. This is where the use of migration data plays an integral role. It’s difficult to argue a net employment increase when people are migrating away. Likewise, it’s difficult to argue a net job loss when inflows are steadily increasing. If NAFTA has in fact had a significant effect on the U.S. labor force, it is reasonable to assume that this will be reflected in the observable migratory patterns. Therefore, migration flow data will be a strong indicator of true impact.
Methodology
In order to determine the migration change inflicted by NAFTA, the regions observed needed to have manufacturing as an industry focus. Therefore, the first step involved identifying regions with the highest manufacturing employment. The Quarterly Census of Employment and Wages (QCEW) data files were used to compile a list of the top 25 U.S. counties with respect to manufacturing employment. From there, the Statistics of Income (SOI) Tax stats – migration data taken from the IRS were used to track county-to-county migration flows. Migration data from the following years was analyzed: 1993, 1994, 2000, 2008, and 2014. 1993 and 1994 were the most indicative of immediate migratory change. They represented the last full year of manufacturing output before NAFTA and the first full year of output post NAFTA. Six years later in 2000, the data was predicted to show migratory patterns of a nation well-adjusted to the trade deal. The data from 2008 was important to show any possible impact that may have resulted from the recession, and 2014 was the most current data available. Analysis involved recording total inflows and outflows for each of the 25 counties in each of the 5 years. Once recorded, a line graph was constructed for each of the counties showing the change in inflow and outflow levels over time. The average inflow and outflow for all 25 counties was taken for each year observed and compiled into a single scatter plot with trend lines indicating the pattern of inflow and outflow for these regions as a whole since 1993.
Results
In order to support the claim that NAFTA has resulted in net job loss, the data was anticipated to display increasing outflow for the observed regions, which were undoubtedly the most affected. On a line graph, this would require the line for outflows to have a positive slope and the line for inflows to either be negative or if positive, have a lower slope than that of outflows. Referencing tables 1-25, which represent the inflows and outflows of all 25 observed counties, it is apparent that the hypothesis was not only unsupported, but directly opposed. The majority of counties showed increased inflow rates and decreased outflow rates. Table 26, showing the trendline for inflows and outflows on average across all 25 counties, makes it clear to see that overall, migration to these regions is on the rise.
Tables 6 and 16 show particularly clear visible representations of the increasing rate of inflows and decreasing outflows with a positive vs. negative slope. However, the most common case is a county with initially high outflows and low inflows that over time shows a closing deficit whether the slopes are positive or negative. This can be observed in tables 1, 2, 3, 4, 7, 10, 11, 12, 15, 17, 18, 19, and 23.
At this point, the migration data displayed massive inconsistency with the job loss claims. How could counties with the most manufacturing employment be growing if the jobs were being shipped across the border? It doesn’t line up. In order to support the migration data’s suggestion that jobs aren’t in fact being lost, but rather being created, it was necessary to return to the QCEW data files, and see if the employment levels were in fact increasing in these counties during the years observed.
Refencing tables 27-51 which show rate of change in employment across all industries for each of the 25 counties, 24 out of 25 counties showed net job loss within the manufacturing industry. Most of which showed quite a significant drop (New York County -73.98% & Cook County -51.80%). If these were the only jobs observed, that would leave a huge question mark as to how these counties could possibly experience increased migration inflows. Fortunately, the gap is accounted for in an overall net increase of jobs. 19 out of 25 observed counties showed a positive rate of change for employment when all industries were factored in. This means, although manufacturing is on the decline, we aren’t really sending any jobs away. Rather, they are shifting in industry concentration. Table 52, which shows the average rate of changes in employment for all the counties in each industry shows exactly where the jobs have gone.
The average decline in manufacturing across the 25 counties was -25.14% (the lowest of all industries). In contrast, the average change in overall employment across industries was 20.68%. The growth trends show the strongest increases coming from the education & health services (60.72%), professional & business services (52.16%), and information (39.01%) industries.
The results of this study shift the question at hand altogether. The analysis of migratory patterns indicates that the loss of jobs may in fact not be in question, but the source of jobs seems to be.
Conclusion
The combination of migration data with the QCEW employment files suggests the hypothesis is not true and jobs are not being “lost” but rather transitioning industries. Sure, manufacturing employment has declined, but those jobs are being replaced with new jobs in the same regions, only in different sectors.
This could add another perspective to the conversation about trade and the American worker. Undoubtedly jobs in the education and health industries require higher levels of skill and education. Therefore, the transition of jobs to this industry requires lots of job restructuring and training. As such, my suggestion for further research would be to perform an analysis on the cost of employment shift into higher skilled industries.
Areas for possible refinement in the research include a broadened and more standardized timeline as well as a scaled version of the county level data to include state and national data. Adding more than 5 years of observations allows us to have a more comprehensive view of the data, and standardizing intervals would improve the accuracy of growth rates.
The research did face a notable limitation. The use of nominal employment figures as opposed to manufacturing as a percentage of total employment was used to determine the counties observed. Although using manufacturing concentration may have showed greater impact than highest manufacturing employment numbers, it would be a poor reflection on the U.S. as a whole. Many counties with the highest industry concentration are marginally smaller and represent only a few thousand people. The use of larger counties was an attempt to scale the national impact more accurately.
Since the implementation of the North American Free Trade Association (NAFTA) in 1994, increasing awareness of the widening trade deficit in combination with claims by the current administration have suggested that the United States has sent many jobs across the border to Mexico. In that case, the loss of jobs would undoubtedly impact the migratory patterns for regions whose industry concentration is manufacturing. This begs the questions, how have migration flows been affected since NAFTA was enacted and what does this reveal about overall employment? If jobs are in fact being lost to Mexico, it is reasonable to hypothesize that regions with the most manufacturing employment would see decreased inflow rates and increased outflow rates.
Theoretical Basis
There’s been countless studies published within the last decade attempting to assess the true impact that NAFTA has had on the economy. Critics of the deal, including the current head of administration, argue that, “it stripped us of manufacturing jobs. We lost our jobs. We lost our money. We lost our plants. It is a disaster.” (Trump - Oct. 9 in St. Louis, Mo.) According to an article published by the Federal Reserve Bank of Dallas, nearly 710,000 jobs were lost between 1994 and 2014 as a result of increased imports from Mexico and Canada or due to shifts in production. According to Brooks Jackson and FactsCheck.org, anti-NAFTA politicians advertise the loss of 1 million jobs. However, contrasting studies have concluded the trade deal resulted in much smaller job losses or even a small net gain.
NAFTA supporters highlight the creation of jobs since 1994, saying that the deal has, in reality, raised employment. In fact, the U.S. Chamber of Commerce claims that NAFTA has brought millions of jobs to the U.S.; be that as it may, those claims count only the jobs gained by increased exports. This neglects the jobs lost by increased imports.
Taking both of these arguments into account, it’s still difficult to assess the true impact that NAFTA has had. Naturally, in any given industry, there will be the elimination and creation of jobs over time, but neither of those figures are useful indications of the strengths and weaknesses of NAFTA when not looked at in context. The full story can only be told with both. Additionally, it’s hard to trust the objective nature of data when the source is often biased in one way or another. This is where the use of migration data plays an integral role. It’s difficult to argue a net employment increase when people are migrating away. Likewise, it’s difficult to argue a net job loss when inflows are steadily increasing. If NAFTA has in fact had a significant effect on the U.S. labor force, it is reasonable to assume that this will be reflected in the observable migratory patterns. Therefore, migration flow data will be a strong indicator of true impact.
Methodology
In order to determine the migration change inflicted by NAFTA, the regions observed needed to have manufacturing as an industry focus. Therefore, the first step involved identifying regions with the highest manufacturing employment. The Quarterly Census of Employment and Wages (QCEW) data files were used to compile a list of the top 25 U.S. counties with respect to manufacturing employment. From there, the Statistics of Income (SOI) Tax stats – migration data taken from the IRS were used to track county-to-county migration flows. Migration data from the following years was analyzed: 1993, 1994, 2000, 2008, and 2014. 1993 and 1994 were the most indicative of immediate migratory change. They represented the last full year of manufacturing output before NAFTA and the first full year of output post NAFTA. Six years later in 2000, the data was predicted to show migratory patterns of a nation well-adjusted to the trade deal. The data from 2008 was important to show any possible impact that may have resulted from the recession, and 2014 was the most current data available. Analysis involved recording total inflows and outflows for each of the 25 counties in each of the 5 years. Once recorded, a line graph was constructed for each of the counties showing the change in inflow and outflow levels over time. The average inflow and outflow for all 25 counties was taken for each year observed and compiled into a single scatter plot with trend lines indicating the pattern of inflow and outflow for these regions as a whole since 1993.
Results
In order to support the claim that NAFTA has resulted in net job loss, the data was anticipated to display increasing outflow for the observed regions, which were undoubtedly the most affected. On a line graph, this would require the line for outflows to have a positive slope and the line for inflows to either be negative or if positive, have a lower slope than that of outflows. Referencing tables 1-25, which represent the inflows and outflows of all 25 observed counties, it is apparent that the hypothesis was not only unsupported, but directly opposed. The majority of counties showed increased inflow rates and decreased outflow rates. Table 26, showing the trendline for inflows and outflows on average across all 25 counties, makes it clear to see that overall, migration to these regions is on the rise.
Tables 6 and 16 show particularly clear visible representations of the increasing rate of inflows and decreasing outflows with a positive vs. negative slope. However, the most common case is a county with initially high outflows and low inflows that over time shows a closing deficit whether the slopes are positive or negative. This can be observed in tables 1, 2, 3, 4, 7, 10, 11, 12, 15, 17, 18, 19, and 23.
At this point, the migration data displayed massive inconsistency with the job loss claims. How could counties with the most manufacturing employment be growing if the jobs were being shipped across the border? It doesn’t line up. In order to support the migration data’s suggestion that jobs aren’t in fact being lost, but rather being created, it was necessary to return to the QCEW data files, and see if the employment levels were in fact increasing in these counties during the years observed.
Refencing tables 27-51 which show rate of change in employment across all industries for each of the 25 counties, 24 out of 25 counties showed net job loss within the manufacturing industry. Most of which showed quite a significant drop (New York County -73.98% & Cook County -51.80%). If these were the only jobs observed, that would leave a huge question mark as to how these counties could possibly experience increased migration inflows. Fortunately, the gap is accounted for in an overall net increase of jobs. 19 out of 25 observed counties showed a positive rate of change for employment when all industries were factored in. This means, although manufacturing is on the decline, we aren’t really sending any jobs away. Rather, they are shifting in industry concentration. Table 52, which shows the average rate of changes in employment for all the counties in each industry shows exactly where the jobs have gone.
The average decline in manufacturing across the 25 counties was -25.14% (the lowest of all industries). In contrast, the average change in overall employment across industries was 20.68%. The growth trends show the strongest increases coming from the education & health services (60.72%), professional & business services (52.16%), and information (39.01%) industries.
The results of this study shift the question at hand altogether. The analysis of migratory patterns indicates that the loss of jobs may in fact not be in question, but the source of jobs seems to be.
Conclusion
The combination of migration data with the QCEW employment files suggests the hypothesis is not true and jobs are not being “lost” but rather transitioning industries. Sure, manufacturing employment has declined, but those jobs are being replaced with new jobs in the same regions, only in different sectors.
This could add another perspective to the conversation about trade and the American worker. Undoubtedly jobs in the education and health industries require higher levels of skill and education. Therefore, the transition of jobs to this industry requires lots of job restructuring and training. As such, my suggestion for further research would be to perform an analysis on the cost of employment shift into higher skilled industries.
Areas for possible refinement in the research include a broadened and more standardized timeline as well as a scaled version of the county level data to include state and national data. Adding more than 5 years of observations allows us to have a more comprehensive view of the data, and standardizing intervals would improve the accuracy of growth rates.
The research did face a notable limitation. The use of nominal employment figures as opposed to manufacturing as a percentage of total employment was used to determine the counties observed. Although using manufacturing concentration may have showed greater impact than highest manufacturing employment numbers, it would be a poor reflection on the U.S. as a whole. Many counties with the highest industry concentration are marginally smaller and represent only a few thousand people. The use of larger counties was an attempt to scale the national impact more accurately.