The death of full employment
About 5 years ago a huge hype cycle stated where people were increasingly talking about the potential to replace jobs with AI. It took a while for this to go really mainstream but eventually it did. When ChatGPT was released to the public a massive shift started to happen. For many years in the West jobs have been replaced by automation, this is partly due to automation improving but also substantially due to the great increase in difficulty and cost in hiring employees in the West. Geographical and regulatory arbitrage leads to situations where it ends up being vastly more expensive to hire people for certain jobs based on where they are working as compared with other locations. Not all of this is of a regulatory nature, for example countries with large coastlines have a natural advantage in industries like ship building compared to landlocked ones. But there's clearly regulatory issues that make some industries gravitate towards certain places.
I started writing this article three years ago when I came across another one of those typical posts that people make where they say that AI is going to shift employment around without any job losses overall. The claim I saw from one commentator was that these new developments will just be like the steam engine where people will find new jobs afterwards and job losses won't be a result. This post was made by an assistant professor of computer science and I ended up having a bit of a discussion with them about what evidence they had to back this claim. Unfortunately they had no evidence at all but to give credit where credit is due they had the intellectual honesty to admit that their statement was just a wild assumption backed by nothing. Interestingly this assistant professor works in the sector that is increasingly in the business of selling credentials rather than offering education. And those credentials, along with the growing cost in both time and money to get them, are a large contributor to changing jobs being far more expensive than it used to be. If switching costs are higher than the benefit people will get from switching careers you'll see an increase in overall unemployment when jobs are lost due to automation (or any other reason). All other things being equal the more expensive it is to retrain the less people will retrain. And that dynamic doesn't even take into account a whole host of other important aspects in the broader job market.
I see this assumption that people can just easily find other employment is frequently held by people without question. This view is pushed heavily by those with special interests in pushing that narrative. One of the biggest culprits in this is the "education" industry, which has a vested interest in selling courses, and in the more nefarious cases credentials, to allow people to switch into other careers. Unfortunately this assumption misses the mark in a few very important ways and the implications for this in the broader economy are very big. I'll write about the political and social considerations of this shift in another post about how the Overton Window is shifting substantially on attitudes towards work and towards unemployment. As a society gets more advanced it takes more time for people to be able to acquire the skills required to do the more advanced jobs that maintain that society. There's a rather large number of highly specialized jobs that modern society depends on that the average person doesn't even know exists. These are often jobs in the supply chain and business-to-business roles that aren't public facing.
If you've been reading this blog lately you'll notice that recently I've been writing about some macroeconomics topics lately. Perhaps the biggest "elephant in the room" economic issue is the growing number of unemployed people combined with very high and very persistent inflation. In the past typically if unemployment spiked it greatly reduced inflation as large amounts of currency stopped circulating in the economy as jobs were destroyed. However things are structurally a bit different in the west right now, monetary velocity is by traditional metrics low but yet inflation remains stubbornly high. On top of all this, for quite some time now the labor force participation rates have been dropping. Many people are not spending their time working in full time jobs, but we now have all sorts of reasons now that we classify someone not working as not contributing the to the unemployment number. For example if someone is "disabled" and not working, then they aren't unemployed in many official stats. When you combine decreasing employment with a decrease in the labor participation rate you quickly see that a large percentage of the citizens are not in paid work at the moment which has profound political and economic implications. Essentially the percentage of people who are employed is far lower than most people realize. The reason politics is so important to understand in this context is that unemployment is a political problem, as such many politicians will apply pressure to reduce unemployment. Low unemployment is often seen in a very bipartisan way as a proxy for good governance so there's extreme pressures to be perceived to be doing something about this. It turns out falsifying the numbers is far easier than improving the employment numbers so there's always a temptation present to do so. A few years ago I started to notice reporting on unemployment statistics being downplayed. In the last 3 years many of these stats are just falsified for political ends. The mechanisms for this manipulation would take a blog post of its own to get into.
In recent times employment numbers have been increasingly cooked and manipulated to serve various political agendas. As a result the full gravity of the underemployment crisis that existed long before the covid pandemic has been harder to see than it otherwise should be. This crisis was accelerated greatly by misguided pandemic responses, the economic fallouts of which are only fully being felt now. Making this crisis harder to see is no accident. But things are now at a point where the real employment numbers are just so low that it has become impossible for anyone who has any contact with other people to take the reported numbers at face value in many countries. I know multiple people who aren't employed right now and everyone else I know also knows multiple people who are unemployed. Some of those people who are not working would not be classified as unemployed, others are just woefully underemployed working casual jobs with insufficient hours of employment. While the transfer payments were running hot in 2020-2021 a common knowledge shift occurred because people had the luxury to entertain unemployment while getting paid benefits to not work. For that period of time people seemed less concerned about unemployment. But now with rampant inflation due to excessively loose monetary and fiscal policy many people are now broke again and are finding themselves in far more dire financial situations that are pushing them towards getting more employment to make ends meet. This doesn't even touch the topic of underemployment which is absolutely huge too.
The major change that's occurred especially in developed countries is that it has started to get cheaper to expend capital to meet business needs than it is to hire local labor. The growth in the number of areas where you can substitute capital for labor is one of the most significant societal changes that has happened in human history but yet gets rather minimal discussion considering that it is one of the strongest undercurrents of societal changes in recent years.
So we have this almost contradictory set of assumptions existing at the same time, we have more people who's jobs are being automated but yet we have a policy framework that still talks about the mandate of full employment.
Technology and employment
Technology is in a phase where jobs are being destroyed by technological advances at such a pace that it is hard for people who have had their jobs destroyed to retrain. More jobs are being destroyed by technology than jobs are being created in some parts of the world. But even in the parts of the world where technology is creating more jobs than it is destroying the process of training up to be productive in those new jobs is not necessarily straightforward. Part of this relates to the more transient nature of employment, the average time people stay in the same job has dropped a lot and companies are far less willing to invest in training people. So what does a company do when there's a new technology that requires some training in order to use?
When I was in Canada I remember people talking about this at work around ten years ago. There was talk about putting together funding for people to retrain when their jobs were destroyed by automation. Many of these discussions came about because of the rollout of self service checkouts that was happening in Canada at the time. The introduction of these was a prominent example and perhaps more importantly a public facing example of technology eliminating a large number of jobs. The staff at the stores that introduced automated checkouts weren't reassigned other duties, they were just fired. Someone working an entry level job like this could go work in some other entry level job without much training but that of course presupposes that there's an employer that's willing to hire them. Given high costs of living coupled with high minimum wages some employers simply can't afford to hire as many staff as they would like, especially if investing in training those staff is required to make it worthwhile to hire them.
Some people had the foresight back then to ask, "what will happen to our society if we eliminate a large number of jobs with technology?". In the past the general approach to this question has sadly been to ignore it perhaps by referencing some sort of tenuous claim to how "efficient markets" will reallocate jobs. But if you dug into these assertions a bit and questioned people you'd find surprisingly few answers. Since then the job market has got considerably less efficient. When you start to actually ask that question you quickly realize there's no easy answers but you also realize that the question has to be addressed as it is one of the most important questions of our era. For starters we have to do away with the assumption that people can just find other jobs easily, we haven't had efficient markets for a while and even with the dramatic rise in bullshit jobs - meaning jobs that could easily be eliminated without an associated economic cost to those organizations - we still have a declining labor force participation rate. I think we are starting to see some very negative impacts from this trend at a societal level already, especially when we look at youth unemployment levels.
A very big challenge is cutting through unhelpful mental models people have of the workforce. Specifically many people view people's employment state as a primarily moral question rather than an economic one. I understand why this thought pattern is prevalent, there's something very important to be said about the opportunities for self improvement that are presented to people from gainful employment in meaningful work. However we should confuse moral thinking with economic thinking no matter how tempting that can be to do.
There's a rather large number of benefits from getting people into jobs, especially people who haven't had a chance to work in a career. I remembered at the time asking the question "what about young people who haven't been working long enough to have had an established career in an area if that area is destroyed by automation?". We report numbers of youth unemployment and general unemployment differently. Beyond this younger people were, and still are, seen to be in a different moral category of people regarding employment.
A common theme since 2000 from a particular left leaning viewpoint was that the school system was seen as being a rock solid option for getting young. It didn't matter if people were enrolling in degrees that had no job prospects outside the university sector, "more education" was the assertion shouted at young people. This divergence has grown over time because much of the educational system has come from a background of getting people ready for the jobs of 50-100 years ago rather than the jobs that will exist 10-50 years from now. Questioning this system is politically risky because there's a number of entrenched interests supporting the status quo, the book "Charter Schools and their enemies" has a good discussion of the dynamics. Putting the political positions on education aside for a moment there's a very clear take away from this book that questioning the educational system is very dangerous right now. Sowell provides a strong argument, backed by actual research, that one of the most important things we can do as a society is to make quality education available to kids. To impart change I think we have to start asking some uncomfortable questions about the status quo because it is clear that students are not getting the education they need in order to be able to do the jobs of the future. Keeping youth in school longer than necessary is very similar in many regards to imprisonment, since the power of the state is literally wielded to keep kids at school against their will in many parts of the world. Unfortunately if we don't improve the educational system we will be creating a new generation of people who are getting prepared for the exact jobs that technology is rendering obsolete, or worse are not getting prepared for anything at all other than perhaps a life of being a cynical activist demanding that other people pay for them. To improve this will likely be not just an educational matter but one of aligning interests, a process that I expect will be a tough one.
What I noticed in Canadian policy around 2010 was that if someone young was seen to train up for a job that was destroyed by automation it was seen as different to someone older who'd been working a while who had their job destroyed. The cultural thinking at that time was that investing in retraining people should be spent on those deemed "worthy" of it. Which at that time was seen as a proxy for how likely they were to be able to provide a positive return to society on that investment in their training. People who had worked for some time were seen to be more likely to be able to return value on their retraining. One the one hand there's a certain appeal to this Bayesian reasoning about the estimation of the return on investment for investing in people who had already had proven careers. There's a downside though, as people get older they have less time to return that investment to society. In any case there's likely a lot of worth in helping people who are hungry for success to retrain, regardless of age, as these people are overwhelmingly the largest economic contributors to real economic output.
I think the thought process was that say you had someone working a few decades in an auto manufacturing plant and their job gets destroyed by automation then you know from their track record that they are willing to work. On a policy level this is seen as being relevant because if you start from the premise that resources are limited for retraining then you need to pick who to re-train and who to invest in. Canada in recent times very much had an implicit zero-sum mentality regarding resources, I'm not sure if this is a line of thinking that the high latitudes predisposes people to or was more an artefact of the retreat of the auto industry in some areas that created an environment of pervasive economic pessimism. The line of thinking goes that the worker who's been working for a few decades is "proven" and therefore is a better risk-adjusted recipient for taxpayer money compared to someone who is less "proven" in the workplace.
There may be some merits to this but there's also a massive downside that's not talked about so much. The downside is that this tends to skew government support towards an older age group for training and a country over the longer term will get less returns for this due to the shorter amount of time the older person can employ the skills that they are getting subsidized to learn. Not investing in the young people of a country is setting up a country for certain future decline. Retraining vs training is a complex situation and the tradeoffs involved are especially difficult to deal with. Unfortunately many people will focus more on moralizing rather than providing effective training and work outcomes. This sort of moralizing comes in many forms and is extremely easy to fall into. Identity politics notions of worthiness are also a large factor in the current political climate as people are finding all sorts of extra reasons to be discriminatory when deciding who to give training to based on traits that those people possess.
I think the main driver in the training vs retraining issue is that most of the "training" young people are getting is beyond useless. Many degrees now are a net-negative for people taking them. They waste multiple critical years of their life to get massively in debt to get a credential that won't give them jobs. I suspect if we removed toxic requirements like having an undergraduate degree to get jobs that don't need those degrees we could remove much of the expensive time wasting that people spend at universities. Freeing up those resources would make retraining look far more appealing. It would also allow us to invest more in the people training for the first time who will end up using those skills that have been so expensively taught to them.
Job elimination and salary compression
So far the number of jobs eliminated by technology outright has been rather small relative to the number of jobs created by technology. But a far more impactful change has been happening where portions of jobs get automated, because many people have a managerial philosophy that anything less than 100% utilization of staff time is a fate worse than all others we see some pretty big changes in the workplace. Technology however has had a huge impact on what jobs look like.
If you look at things as an outsider imagine everyone is working 8 hours a day 5 days a week, then someone finds a way to automate part of that job such that 5 days of work could be done in 4 days. You might be thinking "wow that's great, now people can go to work for 4 days instead of 5". But yet that's not what has been happening, instead the overwhelmingly more common situation for companies is to go from employing 5 staff members to 4 and just pocketing the extra wages that would have been spent. Now wealth inequality goes up and unemployment goes up. While it might be more "efficient" in some senses there's some very real questions we as a society need to be asking about what we should be doing in these situations. If the 5th employee who's now laid off gets another economically and socially productive job we will see a rise in standards of living. But what if the 5th employee is now permanently unemployed as a result? This is the question that we see at the population level when we look at the entire job market. The promise of technology, and the entire appeal of making it my career, was about improving efficiency. However I see in our current political and economic systems that efficiency isn't actually improving the lives of people anywhere near as much as it could be. This is a travesty.
This process could be called job compression, which identifies a way which technology can destroy jobs. This is an expression of how changes can "compress" jobs, say someone makes a highly successful app that makes various companies need to hire less hours of labor to get their jobs done. If the software company can spend one hour of effort and save 10 hours of effort elsewhere this is a 10 to 1 job compression ratio.
In a more practical sense this means less people are employed to get the same amount of short term job output. Longer term this doesn't necessarily hold since having fewer staff can constrain the ability to maintain that productivity. For example a company that hires 3 people to get the same output as a company that hires 1 has more fragility with regards to employees as there's more risk on the company that only has one person to do the job. The other longer term aspect is that training new people in house is harder to do when you have fewer employees since each employee has less time.
Full employment is dead
There's a lot of implications of this that need to be talked about. People don't like to admit that wages are set far more by the cost of replacement of the workers than the value that those workers bring.
There's two main ways in which the cost of replacement can go up in line with inflation, labor can organize to put a price floor on employing people in an industry. In practice when this is done it is usually via unionization. The other way is that the supply of labor starts to get scarce so more money can be demanded by workers for their services. Only in the case where there's very scarce talent for economically valuable tasks do the wages start to adjust upwards towards the upper bound that is formed by the productive output of those workers without a lot of external pressures.
In most other cases wages are adjusting downwards to the lower bound of what wages replacements will work for.
I suspect from this point onwards that the equilibrium of the current state of our admittedly completely corrupt and non-free markets is that full employment is no longer going to be possible going forward. Because we have a market with huge amounts of interventions we will not see unemployment drop beyond a certain level, this will happen just because of the lack of market efficiency. Technology destroying jobs makes this problem worse, however the claim I'm making is additional to this and is subtly different. Technology is now creating a situation where even if there was a completely efficient market we might see some people be permanently unemployed.
When you have significant technological unemployment you'll start to have a massive glut of potential workers at the lower bound. This means that the supply for entry level jobs will get much larger putting downwards pressure on the wages for those positions. Inflation driven by wage growth is not possible with a significantly large proportion of the population unemployed at any point in time.
But more importantly when people don't have work they can do lots of bad things happen at a societal level. Men require some form of work in order to be psychologically healthy. Women are naturally hypergamous in their dating preferences and generally speaking rather strongly prefer that their potential husbands make more money than them. Structural unemployment will therefore directly lead to declines in the formation of families and this will reduce the population growth rate. Lack of family formation will form an extremely strong headwind to economic growth and to population growth.
Revisiting the Phillips curve
The Phillips Curve is an economic model that draws a connection between wages and unemployment. The idea here is that there's a correlation between reduced unemployment with increasing wages in an economy.
A few things that have to be addressed here is that inflation will increase the money supply and therefore will increase wages all other things being equal.
In a sound money system the only way that the money supply can increase is from an increase in GDP. We do not live in a time of sound money however. Looking at just about even inflationary monetary crisis worldwide clear shows that unbounded increases in the money supply actually lead to less employment.
What I want to draw attention to however is that if changes in technology mean that it is cheaper to spend money on technology than staff we will have a situation where unemployment increases independently of the wage level or the monetary base growth.
In mathematics we look at the analysis of functions by considering the nature of the function and also the values for which that function is valid. Far too often people in policy making positions look at functions without analyzing what range of values these functions are valid for. This is not a small error, when you extrapolate functions beyond the range of values for which they are valid you start to see extremely bad errors arise.
A very high profile case of this came up in the modelling of the spread of the Covid pandemic, many people were trying to fit data to an exponential model. This model cannot be true after a period of time because the population that allows for the spread is finite. This leaves a situation where people's predictions were that a pandemic would never ever stop spreading and that the rate of spread would always increase, when in reality that's not how these things work. But people treated these mathematically formulae that were incorrectly applied with a reverence that wasn't deserved because the models were mathematically unsound.
Policy makers that lack the numeracy skills to understand these models are at high risk of being misled by them.
That's why I think there's a discussion of the Phillips curve that's crucial to have. There's axioms underpinning the Phillips Curve that need to be talked about so we can get a sense of when this connection is actually valid and when it isn't.
Popular non-solutions
A temptation is to keep dropping interest rates to deal with "unemployment", the idea being that by making the economy more and more inefficient we will force people to hire more staff to deal with it. The issue is that making money more cheaply available will not add any jobs if the money can be spent on automation instead of hiring.
Broken window economics.
This post is part 20 of the "MonetaryPolicy" series:
- Finally getting around to publishing some monetary policy articles
- Fast things happen slowly then quickly
- Politics of unproductive debt
- Futures markets lower prices, both in good and bad ways
- Why do stable coins matter
- Why is so much financial advice bullshit
- Bank bail ins
- Where is money created
- Bastiat on legal systems and morality
- Transitory inflation means permanent purchasing power reduction
- Problems with Celsius
- Crypto's Lehman moment
- Crypto crash update May 2022
- The myth of the unbalanced government budget
- The 2006 debasement of NZ coinage
- Demand destruction anecdotes
- Central bank interventions and price discovery
- Luna a modern case of hyperinflation
- Collateral crisis psychology
- The death of full employment *