TL;DR: Half.
In a previous post, I shared an example on how developers may phrase technical topics differently to managers. Readers expressed doubt towards managers that need such “dumbed down” language, to which my response was:
Because not every engineer work at a tech company. They might be in retail, fashion, film production, etc., where there just happened to have a need for software development.
But just how many software developers work in a non-tech industry?
Definitions
Where do we get employment data? In the US, the Bureau of Labor Statistics (BLS) publishes statistics on employments. Specifically, Occupational Employment and Wage Statistics (OES) describes how many jobs each industry hires for every kind of occupation.
How do we define “software jobs”? OES, like other BLS datasets, categorizes occupations by the Standard Occupational Classification (SOC) System. What we commonly refer to as “software jobs” are formally known as “Software and Web Developers, Programmers, and Testers” (SOC code 15–1250). Two caveats:
- Although it isn’t apparent in the name, this category also includes UI designers (SOC code 15–1255).
- Perhaps counter-intuitively, this category excludes data scientists (SOC code 15–2051.00) and database jobs (15–1240). If interested, you can still find data for those occupations on OES.
How do we define “tech industry”? OES employs the North American Industry Classification System (NAICS) to identify companies by their business activities. Let’s consider the following:
- Companies that sells software are referred to as “Software Publishers” (513200). Microsoft, Adobe, and Activision Blizzard are prominent examples in this category.
- Cloud service providers are generally known as “Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services” (518000).
- Interestingly, you won’t find Amazon Web Services (AWS) under 518000; Instead, AWS is considered a 541500, or “Computer Systems Design and Related Services”. In fact, that’s also where big corporations like Google sit.
- For the sake of completeness, let’s also include “Computer and Peripheral Equipment Manufacturing” (334100) in our list. This sector contains producers of consumer electronics (such as Apple) but excludes major conglomerates (such as Sony) and semiconductor businesses (like Broadcom).
Now that we are on the same page regarding the concepts, we can go ahead and work on the data.
Data analysis
Querying the data. We can download data via the OES Query System. Let’s make the following selections:
- Occupation: “Software and Web Developers, Programmers, and Testers”.
- Industries: All industries at the “sector” level, plus the industries mentioned above.
- Datatypes: “Employment”.
Cleaning the data. From each sector, we subtract the employment numbers of the “tech industries” within it. Across the “tech industries”, we sum up the numbers and create an artificial entry, which we conveniently name “tech sector”. To enhance readability of the chart, let’s also aggregate long-tail sectors into an “others” wedge.
Plotted as a donut chart, the data is quite self-explanatory:
Discussion
It’s immediately clear that non-tech companies hired half of software jobs. If you there’s one thing you can take away from this post, that’s it.
Limitations. This little chart took me < 1hr to make. Hence, there are many caveats you need to watch out:
- The data above ignores ~50% of total employments for software jobs. Those missing entries are either addressed to a cross-industry company or a governmental entity. If you would like to include them, you can select the placeholder sector “Cross-industry, Private, Federal, State, and Local Government” on OES Query System.
- A company may operate in multiple industries. That means a company may be associated with multiple NAICS codes. I didn’t dig into the details whether and how OES deduplicated employment numbers across them, but the conclusion should be a sufficiently good ballpark figure already.
- Again, my choices of what counts as “software jobs” are quite arbitrary, and I myself am not entirely happy about it. For one, I don’t know where “data engineers” sit in the SOC system, and I think they totally are software people.
- Also coarse are my choices of “tech industries”. For example, I chose to not use the category of “Computer and Electronic Product Manufacturing”, because it would include manufacturers of navigation systems, which aren’t really what would come into my mind when I think of “tech companies”. (EDIT: I removed the mention of one particular company that I always associated with fishing equipments rather than software. Apparently that hurts feelings.)
If you haven’t, I hope this post helps you realize that a large portion of software developers are not working in pure-tech companies; and if you have, you can refer to this article to support your argument next time the topic comes up over beer with your friends.
