Employment Rising: Computer Software Engineers

Just a note of brightness for the labor market:  U.S. employment of computer software engineers is now about 20% above the pre-bust level. Go forth and prosper.


  1. From where did you get the number for 2010? The CPS web site gives me only the numbers up to 2009.

    BTW, going back into history I notice that the category “Computer software engineers” exists only since 2003. Before 2003, the roughly matching categories were:

    Computer systems analysts and scientists
    Computer programmers

    Comparing 2002/2003 category totals and overall category subdivisions, it is very suggestive to assume that these two categories were replaced by:

    Computer scientists and systems analysts
    Computer programmers
    Computer software engineers
    Computer support specialists

    I consider the difference between “programmer” and “software engineer” to be rather doubtful, and when considering the following transition of categories it is probably best as well as reasonable to look at the totals, as apparently the category “Computer systems analysts and scientists” was split, and possibly the classification was a bit at the expense of “Computer programmers”.

    Computer systems analysts and scientists 1,742
    Computer programmers______________ 605
    Total_____________________________ 2,347

    Computer scientists and systems analysts 722
    Computer programmers______________ 563
    Computer software engineers__________ 758
    Computer support specialists__________ 330
    Total_____________________________ 2,373

    When going with that, I get:

    2000: Total 2,496 (thousands)
    2001: Total 2,456
    2002: Total 2,347
    2003: Total 2,373
    2004: Total 2,402
    2005: Total 2,492
    2006: Total 2,437
    2007: Total 2,590
    2008: Total 2,787
    2009: Total 2,593
    2010: ???

    As I don’t have the data for 2010, it is not clear whether 2008 is an outlier, but end to end it certainly looks not so spectacular when considering the larger profession.

    • Mike Mandel says:

      Answering some of your questions in no apparent order
      –remember that I’m showing data for the three months ending in November for each year, while yours is full year data.
      –I calculated the 2010 data off the raw data, suitably weighted to match the published totals.
      –yes, the lack of long-term growth in computer and mathematical occupations is a point that I’ve been making for a while. See, for example, page 2 of this paper

      • Thanks. Yes, I noticed the ‘Nov’ in your graph. Can you please tell me the precise table that you used? Also I don’t understand what “suitably weighted to match the published totals” means. The CPS tables I used (cpsaat11) give specific numbers for all categories. In the monthly data, cpseea19 comes closest, but there the smallest supercategory is “Computer and mathematical occupations” which is quite a bit more comprehensive, and which also doesn’t seem to carry well over the 2003 redefinition.

        Before 2003, “computer programmer” was under “Technical, sales, and administrative support”. If today’s additional “math” categories can be found as separate categories before 2003, one could do a similar exercise as I did for the computer category. Or we wait for the 2010 annual data …

    • As a postscriptum to my numbers, when looking at the evolution of the “computer professionals” categories since their inception 2003, it seems quite possible that the “computer software engineers” category has grown a bit at the expense of the “programmer” and “scientist/system analyst” categories, perhaps by progressing reclassification with “software engineer” becoming the canonical midpoint designation.

      In the software profession, before and possibly into the 90’s the term “programmer” was commonly used to denote anybody working on (not with) computer software *using a keyboard*, as opposed to a pure “designer” who would deal in concepts and be above having to type in actual code. Since then the “pure design” aspect has been deemphasized, undoubtedly in association with more “fast paced” business environments and escalating complexity making full up front design infeasible, leading to more iterative approaches based more on feedback from incremental changes than handed down master designs. The 2003 category redesign was probably just in recognition of that, albeit a decade or so delayed.

  2. Even using cm’s stats, it looks like it went up 10% from peak to peak, ie 2000 to 2008. I don’t think this will ever be a very good indicator though, as it’s extremely prone to automation and outsourcing. It does raise the interesting thought that most service work is highly discretionary. Up to a certain age, I used to always go to the barbershop to get my hair cut. However, in the last decade, I’ve been probably a dozen times at most, choosing to use a pair of clippers instead most of the time. The same calculation applies for all consumers across a host of services, from restaurant visits to tech purchases that pay for IT service workers. So while we have much more control of our finances than we used to, it seems that a service economy like the US is also inherently more pro-cyclical, ie more prone to booms and busts. People probably overindulge in steakhouses, strippers, and massages during the booms but can easily cut back during a recession like this. It is interesting that debt, particularly in the backward housing sector, is not always as flexible, with homeowners taking on loans they couldn’t afford and banks getting stuck with long-term fixed interest rates that are difficult to hedge, largely cuz of the dumb govt decisions that gave us Fannie/Freddie. Going to variable interest rate loans and removing all the govt subsidies for inherently unstable long-term fixed rates is a big part of the solution here.

    • In regard to the trend of employment numbers, in the aggregate as well as particular fields, you have to consider population growth. Even with productivity changes (growth?) in mind, population size/growth is a reasonable reference point for employment numbers. I didn’t attempt to dig up the numbers, but I suspect when you normalize for US population size and particularly the working-age sector, employment will probably look flat or declining except for really booming job sectors.

      Along similar lines, I don’t want to suggest cherry-picking, but as you take larger aggregates of categories, your data will approach the trends of the total.

      However this was not why I moved to a larger aggregate. The reason was as I stated that there was a discontinuity in the category breakdown, and IMO my (speculative) construction was the smallest category containing “compute software engineers” that had comparable numbers across the break.

    • Still a slight rise adjusted for population growth, as the 2010 census had population growth at 9.7%, meaning it was probably less than 9% up to 2008.

  3. The value of the data is in finding whether it validates or contradicts the anecdotal which we are too often forced to rely upon in making judgments. The software job story is one that seems to have consistently contradicted the anecdotal for a decade, in my experience. The software people on the ground seem not to experience in reality what the numbers are telling and they tend to challenge the numbers.

    I can’t name another situation like that. Manufacturing, construction, health care, etc. all seem to have statistics that tend to track randomly reported experience. Thus for software we are tempted to dig deeper or propose exotic hypotheses. I’m not sure about law jobs, as the negative stories probably gain too much traction for their shock value in an arena many viewed as safe, or finance.

    I must say that I’ve not heard new software graduates complaining of joblessness, but neither do I know any fresh outs, because many kids and parents got the message that it is a treacherous occupation — I’m following with interest a couple that will hit the job market in 2-3 years.

    Too bad that the Internet as a means of off shoring seems not to also provide for exploitation of our own struggling regions.

    • Some figures for the civilian noninstitutional population (16+ years of age):

      1990: 189 million
      2000: 213 million
      2010: 239 million

      That’s an about 11% growth per decade. When you contrast my numbers above for the “computer professionals” employment to that, they don’t look very spectacular.

      When normalizing employment in this touted “growth profession” by population size, you get at best a flat line end to end, and possibly a slight decline. If you take different age subgroups, e.g. between 20 and the nominal retirement age (the latter being a moving target BTW), the picture will probably not change much, and probably not in a favorable direction.
      As I don’t have the employment numbers for 2010, and neither those before 2000, we can do the exercise for 2000-2009 (in thousands). The first table is computer employment vs. population with YOY growth, the second cumulative growth from 2000.

      Year, Comp. Empl, Population, Empl. YOY%, Population YOY%
      2000 2,496 212,577
      2001 2,456 215,093 -1.6025 +1.1835
      2002 2,347 217,570 -4.4381 +1.1515
      2003 2,373 221,168 +1.1077 +1.6537
      2004 2,402 223,357 +1.2220 +0.9897
      2005 2,492 226,082 +3.7468 +1.2200
      2006 2,437 228,815 -2.2070 +1.2088
      2007 2,590 231,867 +6.2782 +1.3338
      2008 2,787 233,788 +7.6061 +0.8284
      2009 2,593 235,801 -6.9608 +0.8610

      Year, Comp. Empl, Population, Empl. cumul%, Population cumul%
      2000 2,496 212,577
      2001 2,456 215,093 -1.6025 +1.1835
      2002 2,347 217,570 -5.9695 +2.3487
      2003 2,373 221,168 -4.9278 +4.0413
      2004 2,402 223,357 -3.7660 +5.0711
      2005 2,492 226,082 -0.1602 +6.3529
      2006 2,437 228,815 -2.3637 +7.6386
      2007 2,590 231,867 +3.7660 +9.0743
      2008 2,787 233,788 +11.6586 +9.9780
      2009 2,593 235,801 +3.8862 +10.9249

      Clearly population has grown monotonically and employment has wiggled. After 9 years population has grown 11%, employment not quite 4%.

      As a bonus, here computer related employment as a percentage of population, with an ASCII art bar graph:

      2000 1.1741 ***********************************
      2001 1.1418 **********************************
      2002 1.0787 ********************************
      2003 1.0729 ********************************
      2004 1.0754 ********************************
      2005 1.1022 *********************************
      2006 1.0650 ********************************
      2007 1.1170 **********************************
      2008 1.1921 ************************************
      2009 1.0996 *********************************

      Does this look like growth to you? We start with the back of the dotcom hump, and have a flat area with a small hump in 2008 (so far). This is the net hiring we got from the Web 2.0/Mobile phenomenon. If Mike’s projection turns out to hold water, we may see another similar uptick in 2010.

      In summary, computer employment was lackluster and mostly kept up with population growth overall. I certainly don’t see a big growth story.

      The experience on the ground (in “Silicon Valley”) matched this – robust growth in Internet/Mobile since a few years, but continued cost cutting and hiring restraint in other sectors. There are no apparent mass layoffs in the larger industry, but widespread stagnation, and offshoring is continuing and seems to be here to stay. Google figures prominently as a destination for people who left in recent years where I work.

      • The tables are quite unreadable in proportional font. When you cut/paste them into a text window with fixed-size font (e.g. Courier), they will look better.

      • Mike Mandel says:

        Hell. I’m having a fight with the tables, for some reason. I haven’t found a conversion process that I’m happy with, including what you suggested. Any other thoughts?

      • Like I said before, software engineering is too dynamic a profession to ever get a good idea from the jobs numbers. There’s just too much scope for automation and outsourcing, the jobs numbers will be all over the place for years to come. However, cm is comparing numbers from an all-time peak in 2000 to the years since, hardly a good measure of overall growth. The real measure of tech’s impact is all the soft-skilled service jobs that increasingly rely on technology, like the Direct TV repair guy who came by yesterday and was just using his smartphone to communicate with work, both to call them and through the built-in 3G internet service. Web designers didn’t exist 20 years ago, now there’s thousands of them. These people aren’t software engineers, but their jobs either wouldn’t exist or would be much less efficient without the pervasive undergirding of tech today. That’s the bigger picture that needs to be looked at, not the small cadre of software engineers that build the tools that are then digitally copied thousands of times over.

      • Ajay: You are right that my starting point is during dotcom. However that is entirely because that’s the earliest comparable data I could readily find. Furthermore I quoted enough numbers so you can evaluate growth from later years.

        I also agree with your point about the bigger picture, and that looking at one pretty circumscribed profession is not the end to all. That was a (minor) part of my point too. I didn’t want to broaden the job category too much, and I’m fairly certain, in a larger category the picture would have indicated similar trends. When you look at the employment/population ratio (i.e. more or less the grand total of all job categories), that speaks a clear enough language – it has been dropping over the whole decade.

        But I stand by my point that in whatever is rightly or wrongly considered the computer profession, there has been no spectacular growth over the past decade when adjusting for population. I cannot make strong points about the future (or even 2010 for that matter), but I doubt we will see strong domestic job growth (adjusted for population) anytime soon in this part of the tech industry, or even in the larger professional sectors.

      • Ajay: On the point of increased labor productivity/flexibility, or at least the potential thereof, (to be) achieved by returns to technology, I remember when I was a kid, or possibly even before that, the promise of then envisioned progressive automation was the leisure society. But imagined in the way that everybody has more leisure, not that an ever shrinking workforce share has more and more work or at least nominally-on-the-job time squeezed out of them while an ever increasing share is surplussed into involuntary idleness but without the wherewithal to enjoy it. Not quite Player Piano but probably close enough.

      • Mike: I reproduced the cumulative table below. I hope I read you right as meaning the graphical presentation of the table. I generated the numbers from the totals of the CPS series categories I stated, and using a Perl script to compute the percentages and the population adjustment, and some manual editing for presentation.

    • Mike Mandel says:

      It’s been a weak decade for software jobs for sure…both the numbers and the anecdotes agree. The question is whether that is finally coming to an end.

      • Well, the decade has come to an end. So will the next one, without any effort from us. 🙂

        As for tech employment and employment in general, I’m not hopeful. For one thing global fungibility of occupations based on general principles and not requiring physical presence in the US, by the nature of the work or enforceable (and being enforced) fiat. This includes everything that has been offshored over the past decades, and then some.

        In locations where these activities are relatively new, the local talent has to grow and catch up to Western (which should rather say industrial) skill/experience levels and work ethics/organization, business conduct, and social interactions matching the new occupations. The same thing has happened in Western industrialization, only that was long before our time. As people anywhere are fundamentally cut from the same cloth, any limitations will not be in the human “material” but in social factors, for example whether societies, or perhaps just industrial/professional subcultures, can convert from a primarily top-down and in cases feudal management paradigm and an emphasis on loyalty and patronage relationships to more empowerment and encouragement of independent individual initiative. (BTW, what I seem to see in the West is, despite a lot of lip service, a return to more top-down and process oriented management. But perhaps that’s just a feature or maturing industries. OTOH that’s where your reported lack of innovation comes in – there is a lot of maturation going around in business.)

      • For another thing, there is the productivity story. IT, and in particular real-time supply chain management and modeling/planning support, has removed a lot of buffers from the system, and enabled paring down workforces, inventory levels, and various other types of buffers or redundancy. This not just in industry. Aside from increased jitter and volatility, in accordance with control system theory (cybernetics), one major result is that the regular business is done with less, and less workforce demand leads to lower pay and inferior working conditions. By way of the consumer economy feedback loop, that leads to price pressures and reduced demand for most things our industries are producing. On the business side, at some point things are good enough and need not be done much better (though they can – but it won’t make a lot of difference in most aspects). Then you go from purchasing new technology paradigms to incremental replacement of broken stuff or the occasional point upgrade. This is not the stuff from which big margins are made.

        A certain percentage of (generally highly productive) industries produce things/services that are needed to sustain the daily workings of society – maybe up to a half. The rest is what amounts to consumer/consumption goods or things done on account of consumption. Iphone etc. are a big success story, but they cannot carry the flagging discretionary consumer economy.

  4. Test whether various HTML tags work.




  5. The <PRE> tag works, so one can display fixed layout. But no tables.

    Year, Comp. Empl, Population, Empl. cumul%, Population cumul%
    2000  2,496  212,577
    2001  2,456  215,093   -1.6025   +1.1835
    2002  2,347  217,570   -5.9695   +2.3487
    2003  2,373  221,168   -4.9278   +4.0413
    2004  2,402  223,357   -3.7660   +5.0711
    2005  2,492  226,082   -0.1602   +6.3529
    2006  2,437  228,815   -2.3637   +7.6386
    2007  2,590  231,867   +3.7660   +9.0743
    2008  2,787  233,788  +11.6586   +9.9780
    2009  2,593  235,801   +3.8862  +10.9249


  1. […] This post was mentioned on Twitter by Toby Elwin. Toby Elwin said: Employment Rising: Computer Software Engineers: Just a note of brightness for the labor market:  U.S. employment… http://bit.ly/fHEqOW […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s


%d bloggers like this: