The Pincer after the North American Programmer’s Job
Are AI and outsourcing coming for us all?
“All predictions are wrong, especially about the future” - Yogi Berra
“History never repeats itself, but it does often rhyme.” - Mark Twain
A pincer movement is a military tactic where two or more forces converge on an enemy target from different directions in a manner that resembles a pincer or a pair of pliers. This tactic aims to trap the enemy between the converging forces and to reduce their ability to retreat or maneuver. I will analyze the pincer movement of outsourcing and AI going after North American software developers' jobs and their high salaries. The terms developer, engineer, and programmer are used interchangeably, as I am talking about anyone who writes code in this article. Also, being a Canadian, I will use the term “North America” to lump us with the United States.
Note that while I present some economic analysis, it is not a rigorous one to keep the material manageable and the content simple. A full analysis would consider statistical labor data and a Ricardian-type comparative advantage analysis. The topic can be developed into a full economics paper (it probably has been), but that's not the objective of this essay.
2000 when the party was over
It was predicted 20 years ago that high wages for North American programmers would end due to outsourcing. The prediction has failed to materialize. Even during the 2008 recession, companies had invested in IT to automate jobs, and thus software engineering remained in high demand and paid well.
Despite the widespread availability of outsourcing since the early 2000s, the dearth of skilled outsourcing resources and the significant demand for developers have ensured that North American developers continue to command high salaries. Labor shortages have also been filled by the immigration of those overseas programmers who were highly skilled, as well as helped to keep those who have obtained North American University degrees.
The H1B Visa in the United States was created to mainly bring the best and brightest programmers from overseas. The same can be said about the skills point-based immigration program in Canada, which during most of the past 20 years gave a lot of points for programmers (the author is a recipient of a point system immigration visa to Canada and is now a Canadian Naturalized citizen).
Pretending to remote work elsewhere
Compounding everything is the necessity of working in the office rather than remotely. As a matter of fact, some of us remember a time in the early 2000s when remote work was all the rage. And then workers were called back into the office. In 2013, for example, Marissa Mayer, the new Yahoo CEO, ordered her workforce back to the office. A wave of corporate directives followed this to return to the office. Flashback much?
The North American management psyche is entrenched in the idea that workers are unproductive unless monitored by middle managers. Elon Musk echoed this sentiment when he said his workers “should pretend to [remote] work somewhere else.”
This déjà vu will bode well for North American job security and high wages if we ignore the quarter of a million technology workers laid off in late 2022 and the beginning of 2023. To explain the latter, the theory goes that big tech has overhired during the pandemic, and since customer demand did not materialize, they are reducing their costs. Here in Montreal, I remember seeing a lot of job postings on LinkedIn for the major tech firms during the past three years, but not so many now.
Remote work seems here to stay. Something is different about post-COVID-19. This time most, if not all, white-collar workers had a taste of working from home. And with the new tools or rather the refined and improved tools we have all used, most of the workers want to maintain it. Furthermore, most bosses have seen good or improved productivity figures during and after the pandemic with remote work, especially for software engineers. And so, despite all the noise from the likes of Elon Musk and tech layoffs, most programming is still taking place in the home offices. And programmers who are noted introverts want to maintain it. And thus, with the recession, the cost-cutting attractiveness of outsourcing is once again taking center stage.
20 years and better skills
And twenty years later, worldwide developer skills have been improving. For instance, taking China as our first example, an article from gaper.io says, “If there were an Olympic competition for programming, China would come in first place, according to the website Hacker Rank.” On the other hand, India boasts many big American outsourcing and technology companies, so the skills from America were transferred there long ago.
As for costs, the salary figures put many smiles on the bean counters’ faces. According to one report, a Chinese developer earns, on average, USD 1457 a month. At the same time, an Indian developer will earn around USD 455. A far cry from the USD 5000 that a Canadian developer makes (which is the lower figure in North American developer salaries).
Outsourcing constitutes one flank of the pincer going after the North American programmer job and high salary. The other flank is the rise of generative AI, automatic code writing tools, and no code/low code generator.
ChatGPT Me
As of this writing, while still not widely adopted outside of big technology companies, tools like GitHub Copilot, and ChatGPT have been reported to achieve development productivity gains of 80%, as noted by the programmer and AI pioneer Andrej Karpathy. Furthermore, various experiments have occurred with the new Low-Code/No-Code tools that enable business people to make some of the software that only programmers could create previously.
If we assume that layoffs happened because technology companies could do the same amount of work with fewer workers (the semantic meaning of the word "redundancy"). One possible culprit is generative AI or anticipating its near-future potential.
Considering the big technology companies that had massive layoffs and assuming that they were motivated by AI productivity gain (a strong assumption, I know), we find that for Alphabet, Microsoft, and Meta and following Table 1, they laid off around 5% of their workforce ( I didn’t include others most notably Amazon as its labor force is far beyond the tech workers and couldn't find exact data. Also, I took for granted that job losses are software engineering jobs, which are just for illustration purposes). This comes down to a productivity gain of 6% for AI tools. Given that there are 4.3 million software developers in North America, and 250 thousand were laid off at the time of the writing, this, too, seems to correlate with our 5% number.
Correlation is not causation, and so far, this is superficial (I admit it); we need further and more advanced statistical analysis to find the correct correlates for the layoff. However, anecdotally, I have been experimenting with AI code-generation tools after work. And I have found a much higher productivity figure than 6% but maybe not 80%. I also worked as an outsource developer years ago; therefore, I attest to the high skills out there, hence my inquiry in this essay.
Will the pincer break flank?
Escaping Malthus’ trap but not into Keynes’ Dream
If we are guided by history, then maybe not. In 1798, a classical essay by economist and philosopher Thomas Malthus argued that population growth would bring about poverty, famine, and environmental problems. Yet that was just about when we saw the huge explosion of population growth and immense wealth. This all came about due to improvements in technology.
In the 1930s, economist John Maynard Keynes wrote an essay titled “The Economic prospect for our grandchildren,” he predicted that by 2030, people would need to work 15 hours a week. In other words, we won’t need to work as much in an era of economic abundance. While we have more abundance, we have more needs, so the work week has decreased since the depression era, but has remained at 40 hours since 1940.
Technology has generated job growth, which has generated abundance and more needs and jobs. It's an economic virtuous cycle. According to the US Bureau of Statistics, the employment of Software Developers, Quality Assurance Analysts, and Testers is projected to grow by 25% by 2031, significantly faster than the average growth rate of 5%. This shows the cycle at work. We need more software developers.
Yes, the big technology companies might be laying off people now (and I believe it’s a big mistake as they have lost the confidence of their workers). But outside of Big Tech, the need is there, as an article in Barron’s indicates. In fact, in early 2023, Hired (a recruitment site) posted that jobs for software engineers and developers are still witnessing growth with high salaries. A similar claim was made in US News in an article titled “Why Software Developer Is the No. 1 Job of 2023.” And this job still commands a higher-than-average salary globally (although more in North America than Western Europe, and more in Western Europe than Eastern Europe, India, and China).
New Opportunities ahead
With that in mind, let us see where the growth in demand for software engineers might come from.
The automotive industry has seen a surge in automation capabilities since Tesla entered the market. This has led to a significant increase in demand for software that can operate both in cars and the physical world, including IoT devices. As a result, the need for skilled software engineers is on the rise. The same can be said for smart grids and renewable energy networks.
The enterprise, which always was a big employer for IT developers and consultants, is embarking on modernization efforts to rein in costs and quality issues associated with its aging legacy systems. Where do you find a 1980s mainframe still working? Look to the heart of your bank and government institution. Work for software engineers is growing with the impetus of Cloud migrations, which is becoming primordial as businesses scale. This touches many companies, from financial institutions, transportation companies, utilities, government services, and the healthcare industry. And the consumer and regulators are more adamant about pushing this migration and retiring legacy software. For instance, the recent grounding of Southwest airlines was a call to arms to upgrade their aging IT system.
The last two paragraphs referred to traditional and deterministic Apps. But AI, the new kid on the block for the first time in a long time, is opening a new world of possibilities for what can be done with consumer, enterprise, and business software. Complicated jobs that require data management or reading and reviewing documents, such as legal documents, are being automated. Bad for wider white-collar jobs. Maybe? (I would argue that there will also be new jobs there). But good for the software business and good for the startup and unicorn business. For example, despite the recession, there is growth in venture capital funds for AI-driven startups.
This drive to AI and automation also eats the more traditional outsourcing market. The new generation of AI chatbots powered by ChatGPT-like functions will revolutionize call center operations.
Work the same job but don’t earn the same money
The high demand still doesn’t explain why there should be a disparity between North American and overseas programmers. In essence, you have two groups of workers, A and B, one that earns a high North American salary of $A and the other a low developing nation salary of $B. But you also have, as I am arguing, a total global demand C that is increasing and exceeds both A and B. A supply and demand analysis should indicate that the result will be a demand curve moving to the right (figure 1), given a stable supply, resulting in a higher price for labor (and slowly more developers entering the job market to increase the supply as well, but so far they can't keep up with the demand). In this case, logic would say that the outsourced developers who are more in demand due to their lower wages and equal skills will capture more of the excess price. Indeed, a recent study claims that there will be a 70% increase in outsourcing.
Economic theory says that North Americans earn more because they are more productive, so it will justify the wages in North America on this basis. If this is still true today, then it explains the disparity. However, comparing Canada and the United States, you will face a more elaborate version of the economic puzzle: Canadian workers are less productive than their American counterparts. This is well: "puzzling" because the two countries are very similar in education levels, culture, history, language (one is in common, and most Canadians speak it), general wealth, and ethnicity. The study claims that there is low innovation in Canada for labor, at least.
But my experience again, as someone who used to work in a low-salary country (Lebanon) and whose salary magically increased by making it to Canada, I, and many others like me, don't buy this explanation. I think salary differences might come from the cost-of-living negotiation and lack of job alternatives in different nations. If the North American programmer salary were to reach the South Asian salary level, many in North America would leave for other higher-paying jobs. And because of the higher GDP in North America, these jobs still exist. In South Asia, there are fewer choices.
A third partial explanation suggests that the salary difference may not be as significant as the previously presented figures suggest. It can be attributed to the additional costs incurred when hiring programmers through outsourcing companies. Directly employing remote workers from India or China is difficult for North American companies. Hence, they resort to outsourcing companies that charge an overhead fee for managing resources in these countries.
The Luddites strike back
With that in mind, I hope business leaders will wise up to this era’s software potential and keep hiring engineers here and abroad. After all, technology is what increases GDP and helps us all escape Malthus' trap. However, the rising demand for software is not guaranteed. Bean counters can choose to cut software costs and decrease investments in technology during a recession (self-defeating, but it increases short-term stock prices for non-technology firms). In neoclassical economics, the demand curve will move to the left (figure 2), with businesses asking for less labor. Prices go down. The higher-priced North American developer leaves the job market. And so, the 4.3 million voters and their families that depend on programming incomes would resort to measures to guarantee job security.
They could widen the moat that currently stops regulated industries from hiring outsourced professionals to include every company in North America. In his “World is flat,” Thomas Friedman recounts a story where an outsourcing company was banned from bidding for a contract because they were not local. Furthermore, the current trade war with China will make outsourcing to that country difficult (less supply). Such measures can make a comeback by reducing profit and international competitiveness, and economics is clear about the benefits of trade, which includes the market for jobs.
Another policy could see some companies banning Large Language Models from being used by developers. Arguments can run the gamut from saying that they increase code bugs (hint do TDD and BDD, which you should do anyway with your code and see how many bugs you get with LLMs) to having code content shipped to the servers of the big tech company that offers the LLMs (for ChatGPT and copilot it would be Microsoft). Such measures will also reduce productivity and affect overall GDP.
Living in North America still has advantages that would help North American developers cope. For instance, developers could heed the advice given to factory workers in the American Rust Belt: "reskill and retool." In North America, we have access to the world’s top educational institutions. Some high-end machine learning jobs pay seven figures per year but require advanced degrees like PhDs. North American entrepreneurs have access to the largest venture capital funds, so going into entrepreneurship will also bring about huge gains.
On the flip side, looking at the outsourced worker, and given a race to the bottom spiral, the nightmarish scenario ending North American programmer careers, would turn the outsource software company into a sweatshop. Over there, developers will be run by factory-type lean Toyota production-style methods and, prompted by LLM models, spending long hours for little return.
Therefore, it is in everyone’s interest and the broader economic interest for corporations to invest in R&D and software (just consider that after the recession, the ones with dominant software will conquer their markets). So, the total supply of developers will continue to be lower than the demand until outsourced developers' salaries rise enough to render the difference moot. We can only hope that the flank of North American programmers will get the proper support from their developing world brethren, and vis versa (let our love of code Unite us).
Epilogue
The pincer movement is likely to be tried again in the coming years. The newest flank is the generation Z cohort of graduates for whom software engineering is the preferred profession. They will soon be flooding the junior developer market. They might already be embracing AI tools and have the right skill set, but they can always ask to return to the office if they face pressure from outsourcing.
Comments
Post a Comment