From national security, to productivity growth and innovation, to qualitative ideas of what a “good job” is, manufacturing holds a special place in the minds of policymakers.
Thanks for researching and presenting a very thoughtful article.
Is it fair to say that part of this issue is how inclusive we make a definition of manufacturing, whereby an administrative person located in the front office of a manufacturing site might consider themselves to be "in manufacturing," even though they do not directly manufacture anything? I think the (overly simplistic) definition of what manufacturing is would be something like, "to take one thing and add it to another thing to produce a third thing." Anyone who isn't doing that with their hands or operating a machine that does that isn't a manufacturing employee, but they may still be engaged in the manufacturing industry. Instead, they are employed in administrative, repair, human resource, or other supporting roles. While it's tempting to modify the definition of "manufacturer" to include these technical, logistics, and other support functions because of the explosive roles of technology and offshoring have had on the industry, I think it would be doing a disservice to the millions of people who are no longer employed as a manufacturer, thereby doing damage to the overall understanding of how this industry has changed over time. (Although maybe not, since all the metrics show the same drops since 1975-1980; if you take out the 2008/9 recession/recovery, it's almost a straight line down and to the right.) To be "in manufacturing" now, I would advise people to consider a degree in one of the support functions.
A long-winded way of saying I think your ultimate conclusion about relying on more than one metric is sound, as is very carefully defining the variables used. Again, great work...thank you!
Great question. Census assigns establishments (physical locations) a single industry code based on its primary activity, which is determined by revenue. That is what flows into the BDS, after all of the longitudinal consistency algorithms to translate across vintages of industry codes. More on that here https://www.census.gov/library/working-papers/2021/adrm/CES-WP-21-08.html
This is a good discussion. One more complication is that company founders first have to classify their startup at the firm level on the SS-4! Then some time later they may receive one of the establishment-level surveys that asks them to classify the establishment, possibly according to a different rubric. There's lots of potential for misunderstanding and maybe even some worries about giving an answer that appears to contradict the one on the earlier form. It appears that the establishment surveys have improved a lot since I filled out one as an entrepreneur in the 1980s. I was a cofounder of a software firm that (pre-Internet) distributed book-like software on floppy disks. I remember being genuinely confused about whether that made our establishment a manufacturing or a software or a publishing operation. I believe there remain many such ambiguities just on the employer side.
Very interesting point. Important to remember that respondents (business owners) don't think in terms of NAICS codes. As Rick notes below, lots of different pieces of information feed into the Census Business Register, including industry codes from the BLS, so there is some hope in triangulation.
Thanks for this article. I worked at BLS for 40 years, the last 17 I ran the QCEW which is the business universe program. The QCEW collects NAICS industry codes for about 12 million business establishments, including a million new businesses each year. We also update those codes for most business establishments on a 3 year cycle to maintain currency. (some businesses like cemetaries and dental offices are unlikely to change and are checked every 6 years). The authors got a lot right, but the key item for policy and analysis is that once the definitions are established (a huge effort) it is the trend that is key. Is something going up or down? Given a consistent process and strong statistical processes for controlling errors, the relevant issue is direction.
One point left out is that by classifying non-manufacturing locations that might be a part of a larger manufacturing business into their specific economic activity, then all of the occupational data, projections and other workforce measures, occupational training needs, etc., are far more accurate.
It is also important to state that the various Census Bureau business programs get the benefit of all of the BLS work to identify the correct industry code and maintain its currency. Each year BLS shares with the Census Bureau millions of industry codes, county codes, locations, addresses and the establishment breakouts for multi-unit firms. This keeps government data from the various agencies as consistent and cost-effective as possible.
Lastly, getting codes and locations and other key information right is critical and expensive. So, every dollar cut from the BLS QCEW hurts the accuracy of both BLS and Census data.
Thanks for your comments. Pulling folks with deep QCEW experience into the Substack comments means we are doing something right!
I would say that both the levels and the trends matter, and as we point out here, different data sources can tell you different types of information. I think an interesting side point here is that these trends (and levels) are about changes in the composition of business activity. In this context it might be the reclassification of continuers (manufacturing plants switching to R&D estabs) or the composition of entrants and exits (net outflow of manufacturing estabs). It turns out that most of it is driven by entry and exit flows rather than reclassification (see Teresa Fort's work on that here
Also, having spent years working with Census microdata, I couldn't agree more about the value of having additional classification information flowing into the Census Business Register. I also agree about the importance of properly funding and staffing that statistical agencies.
Thanks. Yes of course levels , trends and the dynamics of new entrants, deaths, growing and declining firms, by size, geography and industry all are needed to paint the fuller economic picture. I all read the article and get back to you.
Thanks for researching and presenting a very thoughtful article.
Is it fair to say that part of this issue is how inclusive we make a definition of manufacturing, whereby an administrative person located in the front office of a manufacturing site might consider themselves to be "in manufacturing," even though they do not directly manufacture anything? I think the (overly simplistic) definition of what manufacturing is would be something like, "to take one thing and add it to another thing to produce a third thing." Anyone who isn't doing that with their hands or operating a machine that does that isn't a manufacturing employee, but they may still be engaged in the manufacturing industry. Instead, they are employed in administrative, repair, human resource, or other supporting roles. While it's tempting to modify the definition of "manufacturer" to include these technical, logistics, and other support functions because of the explosive roles of technology and offshoring have had on the industry, I think it would be doing a disservice to the millions of people who are no longer employed as a manufacturer, thereby doing damage to the overall understanding of how this industry has changed over time. (Although maybe not, since all the metrics show the same drops since 1975-1980; if you take out the 2008/9 recession/recovery, it's almost a straight line down and to the right.) To be "in manufacturing" now, I would advise people to consider a degree in one of the support functions.
A long-winded way of saying I think your ultimate conclusion about relying on more than one metric is sound, as is very carefully defining the variables used. Again, great work...thank you!
Great detail, I had no idea there were this many sources of job classifications, nonetheless that they differed so greatly.
Does the BDS use *primary* NAICS or does the business list all NAICS they operate in?
Similar to holding multiple jobs, businesses can operate in multiple industries, like the janitors in manufacturing.
Thanks again!
Great question. Census assigns establishments (physical locations) a single industry code based on its primary activity, which is determined by revenue. That is what flows into the BDS, after all of the longitudinal consistency algorithms to translate across vintages of industry codes. More on that here https://www.census.gov/library/working-papers/2021/adrm/CES-WP-21-08.html
Thanks for the details.
This is a good discussion. One more complication is that company founders first have to classify their startup at the firm level on the SS-4! Then some time later they may receive one of the establishment-level surveys that asks them to classify the establishment, possibly according to a different rubric. There's lots of potential for misunderstanding and maybe even some worries about giving an answer that appears to contradict the one on the earlier form. It appears that the establishment surveys have improved a lot since I filled out one as an entrepreneur in the 1980s. I was a cofounder of a software firm that (pre-Internet) distributed book-like software on floppy disks. I remember being genuinely confused about whether that made our establishment a manufacturing or a software or a publishing operation. I believe there remain many such ambiguities just on the employer side.
Very interesting point. Important to remember that respondents (business owners) don't think in terms of NAICS codes. As Rick notes below, lots of different pieces of information feed into the Census Business Register, including industry codes from the BLS, so there is some hope in triangulation.
Thanks for this article. I worked at BLS for 40 years, the last 17 I ran the QCEW which is the business universe program. The QCEW collects NAICS industry codes for about 12 million business establishments, including a million new businesses each year. We also update those codes for most business establishments on a 3 year cycle to maintain currency. (some businesses like cemetaries and dental offices are unlikely to change and are checked every 6 years). The authors got a lot right, but the key item for policy and analysis is that once the definitions are established (a huge effort) it is the trend that is key. Is something going up or down? Given a consistent process and strong statistical processes for controlling errors, the relevant issue is direction.
One point left out is that by classifying non-manufacturing locations that might be a part of a larger manufacturing business into their specific economic activity, then all of the occupational data, projections and other workforce measures, occupational training needs, etc., are far more accurate.
It is also important to state that the various Census Bureau business programs get the benefit of all of the BLS work to identify the correct industry code and maintain its currency. Each year BLS shares with the Census Bureau millions of industry codes, county codes, locations, addresses and the establishment breakouts for multi-unit firms. This keeps government data from the various agencies as consistent and cost-effective as possible.
Lastly, getting codes and locations and other key information right is critical and expensive. So, every dollar cut from the BLS QCEW hurts the accuracy of both BLS and Census data.
Thanks for your comments. Pulling folks with deep QCEW experience into the Substack comments means we are doing something right!
I would say that both the levels and the trends matter, and as we point out here, different data sources can tell you different types of information. I think an interesting side point here is that these trends (and levels) are about changes in the composition of business activity. In this context it might be the reclassification of continuers (manufacturing plants switching to R&D estabs) or the composition of entrants and exits (net outflow of manufacturing estabs). It turns out that most of it is driven by entry and exit flows rather than reclassification (see Teresa Fort's work on that here
https://facultynew.tuck.dartmouth.edu/uploads/teresaFort/files/2019_Bloom_discussion_FORT.pdf).
Also, having spent years working with Census microdata, I couldn't agree more about the value of having additional classification information flowing into the Census Business Register. I also agree about the importance of properly funding and staffing that statistical agencies.
Thanks. Yes of course levels , trends and the dynamics of new entrants, deaths, growing and declining firms, by size, geography and industry all are needed to paint the fuller economic picture. I all read the article and get back to you.