In Pursuit of Job Happiness
Economic models and the perfect role
Why is there unemployment? Why are people stuck in suboptimal jobs? In the 20th century, perhaps motivated by the Great Depression, economists created models to study these issues. These early models were very simplistic, assuming that all participants in a labor market have perfect information and no transaction costs. In such a model, everyone “knows” the best job for them in terms of compensation vs. effort. And every employer “knows” which employees they need and the maximum pay they can give them. Thus the only explanation for unemployment is that the market “fails to clear”, that is the compensation an employer offers is too low for what an employee demands.
The solution, as advocated by John Maynard Keynes and his followers, was for the government to step in and provide more jobs or lower the expected wage. This could be done either via fiscal policy (e.g., the government hires a bunch of workers or suspends payroll taxes) or monetary policy (e.g., the government “prints” money, leading to cheaper capital costs and perhaps inflation which lowers nominal wages). It’s somewhat controversial how well these policies have worked out, but nonetheless they remain a cornerstone of responses to economic downturns, for example the recent $2 trillion stimulus package passed in 2020 in response to the COVID-19 pandemic.
Towards a Better Model
As you might be thinking already, the simplifying assumptions that made these models easy to work with also caused the conclusions drawn from them to be incomplete. Clearly it is not the case that workers or employers have perfect information. A perfect illustration of this is given in the tenth episode of the podcast “Peak Salvation”, in which Philip Su talks about how there are many jobs that may be a good match for a worker, if only they knew about them:
It’s weird to me that opportunities like this exist. Auto-body painting paying more than software engineering back then. Line services requiring no skills and almost no actual work. Cruises where you make money being on them. Am I just cherry-picking these opportunities, or is the world actually full of these types of things, yet the average person has no way of hearing about them? [...] Could it be that we could get many people into better-paying jobs suited to their talents and interests simply by being better at advertising what’s available? [...] The inability to discover amazing job opportunities is a great disservice to us all.
Of course, he was not the first to realize this. In the 1960s several economists started investigating more complex models that more accurately modeled the real behavior of workers. (In particular, these included Peter Diamond, Dale Mortensen, and Christopher Pissarides who won the 2010 Nobel Prize in Economics for this work). They considered that most workers stay in a job quite a long time (years), whereas they spend a relatively shorter time in unemployment (months). And since jobs vary quite substantially in many qualitative factors, workers are motivated to spend some extra time to find a good match.
The ways they look for matches include informal conversations with family, friends, acquaintances, and prior coworkers, as well as more formal methods such as “help wanted” postings and internet job search boards. And it’s rather uncommon to get all job offers at once, indeed more often a worker gets a small number (perhaps just one) at a time and must accept or decline it. This allowed them to use techniques from optimal stopping theory to determine the best behavior in such a situation. This still requires many simplifying assumptions, such as that jobs can be strictly given a numerical “value”. Nonetheless it’s a big improvement over the naive “perfect information” models that preceded this.
Still, there are at least two big issues with job matching models. First, they don’t seem to address cyclical or seasonal changes in unemployment. There’s no built-in reason that job matching should get more difficult based on business cycles or seasons. Second and perhaps more fundamental is that we would expect the advent of the Internet and corresponding improved job matching technology to reduce unemployment by giving workers and employers better tools to make matches. Surely with the plethora of job search sites and easier networking it would be faster than ever to find a good job. Yet we saw huge unemployment spikes in the 2001 and 2007-2009 recessions:
Patterns of Sustainable Specialization and Trade
In a 2011 paper Arnold Kling introduces a more complex model that expands the simple job matching model, which he calls Patterns of Sustainable Specialization and Trade. The basic idea is that before the job matching process can even begin, employees need to figure out what jobs they need, and workers need to figure out what skills they need (these are the “patterns of specialization”). If there is a mismatch then there will inherently be unemployment, regardless of how much work is spent on job matching or stimulus. It can take time to create new patterns, and because these need to be sustainable, taking short-term actions to address it (for example, stimulus) will be ineffective.
This model predicts that increasing productivity might increase unemployment (contrary to traditional models), if that increase is not uniform. This seems to explain the post-Internet unemployment increase we see in the chart above. What happens is that increasing productivity in one sector means fewer jobs are needed, but because it takes time to figure out what new jobs arise from that, in the short run there is increased unemployment. This can take many years (as we saw in the very slow recovery after 2009).
To take a small-scale view, consider you have a job as a software engineer. You identify some routine work that you are spending 20% of your time doing. Like a good (lazy) engineer, you spend some time automating this work. In the short-term you’ve increased your productivity but also reduced your “employment”, you now have an additional 20% of your time free. It might not be immediately obvious what you should spend that time on instead, which could cause others to perceive you as working less (not as much of an issue with remote work, perhaps). But after some time you’ll identify (or be given) more work, and now you’re doing more than ever. It’s likely the new work is higher-level, more interesting, and has more impact than the rote work you replaced. But there is that expensive switching cost in the middle, which can feel bad.
In Pursuit of Job Happiness
As the last paragraph hinted, we can apply these models’ insights to benefit our own careers. We even see a similar progression of overly simplistic thinking in business. In any company there are internal roles that have a lot of parallels with external job markets (indeed, in some cases internal company applicants for roles are treated exactly the same as external ones). Every company wants to have a good fit between its workers and the roles they are in, in order to maximize employee productivity and happiness. Yet often this is viewed through the naive “aggregate demand/aggregate supply” lens of early economic models. There is simply a number of “headcount” required, a hiring target, and the company seeks to equalize the two values.
This simplistic approach doesn’t allow much room for job matching, leading to poor fits and unhappy/unproductive employees. A better strategy might be to apply some matching techniques, and make it easier for employees to move internally. There could be internal “job fairs”, “job search sites”, or just informal networking opportunities to help everyone find a superior role. This can prevent employees from leaving to a different company entirely, which most employers would rather avoid. Yet in practice it’s often so difficult to change jobs internally that leaving becomes the easiest option. Giving people more opportunities to grow and be happy without switching jobs would be better for everyone.
Finally we should consider the lessons of the PSST model. The set of opportunities that officially exist within a company are not necessarily the optimal set that should exist. There are always new areas to explore and new technologies to discover. It might be that the best role for a given employee doesn’t exist yet, and needs to be created. This might mean a different project, team, or job responsibility for that person, but a company can be more successful by allowing this flexibility. As an employee you should be thinking about this while working–what job do you see as your comparative advantage? Does it exist already? If not, can you make an argument that it should?
Many of the most successful engineers I know were able to carve out an entirely new role for themselves that no one had thought of. This could mean becoming an expert in a new domain (e.g., ML, security, UX) that no one at your company has sufficient expertise in. This could mean developing some new technology that no one yet knows should exist. The initial response to Facebook’s announcement of React was rather lukewarm, with most seeming to think that this tech didn’t do anything that wasn’t already possible. Yet now there are hundreds of job postings for “React developer”.
Of course it doesn’t have to be on that big a scale of industry-changing. One person I know identified that their company wasn’t tracking data quality in their revenue metrics, and set up a team to identify and fix data issues. This was a type of work no one at the company even thought of as existing, but they were able to show it was valuable and eventually increased revenue by a significant amount just by improving the data used for making decisions.
You too can find a job that makes you happy and engaged. You just might have to create it.