I wanted to ask about this: "BLS figures suggest that employment prospects for software engineers have already skipped this phase and are now headed into moderate job loss, bringing IT occupations overall into mild decline."
What you're referring to here are BLS' employment projections from 2023, which is technically a different data source from the 2024 BLS OEWS figures that I cited. Their 2024 estimate will be published in August or September.
Why does the BLS offer two different figures, you might ask?
Their recently-released figures from the OEWS are from a survey of employers. This is the most timely survey but it is also somewhat incomplete in that independent contractors will be under-surveyed. This leads them to undercount employment of taxi drivers for instance, since they won't survey most individual Uber/Lyft drivers that make up a substantial share of this occupation.
More comprehensive data about occupations comes from the Census ACS' individual survey later in the year. This is usually a better source, and the one I cite for high-level longer term comparisons, but very small occupations can see considerable variance due to sampling error. Some workers may also categorize themselves differently than their employers.
The BLS EP figures that you cited, as I understand them, do their best to estimate the true size of the occupation based on these sources, but the extra time it takes to both synthesize any disagreements and make a projection also means they are more delayed.
When the 2024 EP figures come out, I suspect they will confirm the trend from the OEWS ones, which fits the story of low hiring with some layoffs in this field. They actually made quite a significant downward revision in their projection for software developers from 2032 to 2033, and I suspect their 2034 projection will be lower still. If it somehow came in higher, my takeaway would be that there has been an increase in indie developers not surveyed by the BLS.
All of the different data sources about where occupations are heading confused me too at first, hope this helps clarify!
For software developers, I see the OEWS profile saying a decline from 1,656,880 to 1,654,440.
On your chart you say IT occupations (census) – where do you get that data?
Might you consider either of:
1. Releasing the table where you've compiled the BLS data by year :)
2. Making graphs showing the last ~10 years of employment in various professions of interest given AI? e.g. I think posting a graph on Twitter showing software developers posting their first ever decline would be popular.
A post that's just a round of up of which jobs might already be in decline due to AI (plus which are maybe surprisingly growing) – focusing on trends in the last few years – would also be very popular I'd guess.
Some to cover: animators, software developers, tech managers, translators, call centre workers, secretaries & admin, writers, plus any others that come up a lot in discussions of AI, plus any others that have shrunk or grown a lot recently, or you think might in the future (e.g. life scientists, counsellors, managers).
One of my plans for this summer (especially if I can get some institutional support, let me know if you have any ideas there) is to build a website that tracks all of this so people can generate their own charts and also see the source data.
I currently have a massive Google sheet that has all of this data with dashboards that help me quickly generate these charts but it would be great to put it up in a more publicly accessible way.
One of the trickiest things is accounting for the shifts in the specific definitions of occupations over time. The occupation that I call "Software developers, programmers, and testers" is actually three separately tracked BLS occupations:
- Computer Programmers
- Software Developers
- Software Quality Assurance Analysts and Testers
There has been some movement between these occupations that makes each of these difficult to track individually over long periods of time, so I aggregate them to have more comparable figures. More details on this methodology and how it bridges the various classification systems that have been used over time can be found here: https://www.2120insights.com/p/how-quickly-have-machines-always#footnote-4-143084600
For "IT occupations", I use a broader category that basically includes every occupation that crosswalks to BLS 15-1000 occupations, plus "Computer Operators" (BLS 43-9011/Census ACS OCC 5800)
Makes sense! Sounds like releasing the data might be a bigger project, but maybe you could tweet some of the most interesting charts soon? Let me know & I would be happy to cross-promote.
Yes! I have another piece about the impact of DOGE on the job market on deck, but am wondering if I should prioritize more interesting findings from the 2024 OEWS data first.
Will think about this a little more tomorrow, thank you for your thoughtful questions and support!
The agricultural graph is wild, it doesn't take anyone to feed the US. (What's the difference here in outsourcing vs machines are just soooo good at farming?)
Just from an anecdotal standpoint, it doesn't make sense to me, that this will hurt Junior SWE. I think there may be an overreaction from companies that don't realize what they need or can't work the prompts.
I feel like people underestimate how much stuff we need/want to automate. The learning curve for SWE is now lower, and more people can vibe their way into solutions. But those people will also have a bigger sense of what the possibilities are and who they need to help them complete those jobs. (SWE.)
With server farms and needs like that... how could IT in the broader sense go down? This boom should add to jobs where people have to service the physical tech.
I do want to write more at length about the future of software engineers because I think it has some wider implications just like self-driving technology. Something interesting that I noticed in the recently released 2024 occupational data was that while developer jobs were down for the first time since 2010, there was still significant growth in computer and information system managers. This was actually the _only_ IT-specific occupation that saw significant growth. There's a story here that's consistent with another prediction I made last year about more people moving into management as technology becomes more powerful: https://www.2120insights.com/i/151437867/managerial-jobs
Thanks, this is great!
I wanted to ask about this: "BLS figures suggest that employment prospects for software engineers have already skipped this phase and are now headed into moderate job loss, bringing IT occupations overall into mild decline."
The forecast here is still for healthy growth, and it seemed like the 2023-2024 figures were more like flat than shrinking? https://www.bls.gov/ooh/computer-and-information-technology/software-developers.htm
It seems like IT as a whole is shrinking, but SWE are still flat or growing (in this data).
You're welcome, and great question!
What you're referring to here are BLS' employment projections from 2023, which is technically a different data source from the 2024 BLS OEWS figures that I cited. Their 2024 estimate will be published in August or September.
Why does the BLS offer two different figures, you might ask?
Their recently-released figures from the OEWS are from a survey of employers. This is the most timely survey but it is also somewhat incomplete in that independent contractors will be under-surveyed. This leads them to undercount employment of taxi drivers for instance, since they won't survey most individual Uber/Lyft drivers that make up a substantial share of this occupation.
More comprehensive data about occupations comes from the Census ACS' individual survey later in the year. This is usually a better source, and the one I cite for high-level longer term comparisons, but very small occupations can see considerable variance due to sampling error. Some workers may also categorize themselves differently than their employers.
The BLS EP figures that you cited, as I understand them, do their best to estimate the true size of the occupation based on these sources, but the extra time it takes to both synthesize any disagreements and make a projection also means they are more delayed.
When the 2024 EP figures come out, I suspect they will confirm the trend from the OEWS ones, which fits the story of low hiring with some layoffs in this field. They actually made quite a significant downward revision in their projection for software developers from 2032 to 2033, and I suspect their 2034 projection will be lower still. If it somehow came in higher, my takeaway would be that there has been an increase in indie developers not surveyed by the BLS.
All of the different data sources about where occupations are heading confused me too at first, hope this helps clarify!
Thank you, great answer!
For software developers, I see the OEWS profile saying a decline from 1,656,880 to 1,654,440.
On your chart you say IT occupations (census) – where do you get that data?
Might you consider either of:
1. Releasing the table where you've compiled the BLS data by year :)
2. Making graphs showing the last ~10 years of employment in various professions of interest given AI? e.g. I think posting a graph on Twitter showing software developers posting their first ever decline would be popular.
A post that's just a round of up of which jobs might already be in decline due to AI (plus which are maybe surprisingly growing) – focusing on trends in the last few years – would also be very popular I'd guess.
Some to cover: animators, software developers, tech managers, translators, call centre workers, secretaries & admin, writers, plus any others that come up a lot in discussions of AI, plus any others that have shrunk or grown a lot recently, or you think might in the future (e.g. life scientists, counsellors, managers).
E.g. this chart but starting from the trough after the dot com boom and updated to 2024.
https://www.dropbox.com/scl/fi/1c9f70vl0tknjh6fapn9v/Screenshot-2025-04-17-at-12.02.45.png?rlkey=yqwsv1e9aj6gku7kqywl1st6w&dl=0
Great suggestions, thank you!
One of my plans for this summer (especially if I can get some institutional support, let me know if you have any ideas there) is to build a website that tracks all of this so people can generate their own charts and also see the source data.
I currently have a massive Google sheet that has all of this data with dashboards that help me quickly generate these charts but it would be great to put it up in a more publicly accessible way.
One of the trickiest things is accounting for the shifts in the specific definitions of occupations over time. The occupation that I call "Software developers, programmers, and testers" is actually three separately tracked BLS occupations:
- Computer Programmers
- Software Developers
- Software Quality Assurance Analysts and Testers
There has been some movement between these occupations that makes each of these difficult to track individually over long periods of time, so I aggregate them to have more comparable figures. More details on this methodology and how it bridges the various classification systems that have been used over time can be found here: https://www.2120insights.com/p/how-quickly-have-machines-always#footnote-4-143084600
For "IT occupations", I use a broader category that basically includes every occupation that crosswalks to BLS 15-1000 occupations, plus "Computer Operators" (BLS 43-9011/Census ACS OCC 5800)
Makes sense! Sounds like releasing the data might be a bigger project, but maybe you could tweet some of the most interesting charts soon? Let me know & I would be happy to cross-promote.
Yes! I have another piece about the impact of DOGE on the job market on deck, but am wondering if I should prioritize more interesting findings from the 2024 OEWS data first.
Will think about this a little more tomorrow, thank you for your thoughtful questions and support!
Another paper that uses a similar methodology to track occupational churn (though slightly less detailed than mine) is this one which I've cited before here: https://www.economicstrategygroup.org/wp-content/uploads/2024/10/Deming-Ong-Summers-AESG-2024.pdf
The art at the top of the page is cursed.
The agricultural graph is wild, it doesn't take anyone to feed the US. (What's the difference here in outsourcing vs machines are just soooo good at farming?)
Just from an anecdotal standpoint, it doesn't make sense to me, that this will hurt Junior SWE. I think there may be an overreaction from companies that don't realize what they need or can't work the prompts.
I feel like people underestimate how much stuff we need/want to automate. The learning curve for SWE is now lower, and more people can vibe their way into solutions. But those people will also have a bigger sense of what the possibilities are and who they need to help them complete those jobs. (SWE.)
With server farms and needs like that... how could IT in the broader sense go down? This boom should add to jobs where people have to service the physical tech.
Haha, cursed art was my intention and the AI nailed it!
With respect to agriculture, efficiency is a much, much bigger story than outsourcing. The US does have a slight agricultural trade deficit, but we did have a consistent surplus until ~2020, at least in terms of dollar value: https://www.ers.usda.gov/data-products/ag-and-food-statistics-charting-the-essentials/agricultural-trade
I do want to write more at length about the future of software engineers because I think it has some wider implications just like self-driving technology. Something interesting that I noticed in the recently released 2024 occupational data was that while developer jobs were down for the first time since 2010, there was still significant growth in computer and information system managers. This was actually the _only_ IT-specific occupation that saw significant growth. There's a story here that's consistent with another prediction I made last year about more people moving into management as technology becomes more powerful: https://www.2120insights.com/i/151437867/managerial-jobs