JW Mason has an interesting discovery among the data. Specifically, it looks like the US data series for wages, normalised for shifts in the composition of jobs, is much less cyclical than the raw data. In other words, the business cycle seems to affect wages through composition shifts. In recessions, people lose jobs and eventually get hired back into ones with lower productivity and pay than they had before. People who manage to stick to their jobs through the crisis don’t see much difference. In booms, people who lose their jobs (or quit) tend to get hired into ones with higher productivity, and pay, than they had before.
This makes sense. Imagine that people try to pick a job that suits them – or in economicspeak, that maximises their labour productivity. Imagine also that firms try to hire people who suit their requirements. I doubt this will be very difficult. This is a pretty basic market setup, matching workers and vacancies. Now, consider that people tend to acquire skills and knowledge as they work. This might be something exciting, or it might be as dull as someone in sales building up a contacts book. As a result, people will tend to get onto some sort of career path, picking a speciality and getting better at it.
This might be horizontal – people with a highly transferable skill who move across industries – but I think it’s more likely to be vertical. As they gain in skill, knowledge, or just insidership, they are likely to get paid more. We should at least consider that this matches higher productivity. But then, there’s an explosion – suddenly a lot of firms fail, and their employees are on the dole. They now need to search their way back into work. It is likely, at least, that if they have to find it in another sector or even another firm they will lose some of the human capital they acquired in the past. The unemployed are suddenly driven off their optimal productivity path, and are usually under pressure to take any job that comes along, no matter how suboptimal. Until they get back to where they were before the crisis, on their new paths or on their old ones, the economy will forego the difference between their potential and actual production. You could call it an output gap, but that’s taken, so let’s call it the snakes-and-ladders model.
This, in itself, is enough to explain why unemployment is a thing – you can’t price yourself back into a job with a firm that has gone bust – and why productivity might be depressed for some time post-crisis. In the long term people will climb the ladder again, but this is deceptive. Society, and even firms, can think of a long term. Individuals cannot, as life is short. As the man said – in the long run, we are all dead. Hanging around at reduced productivity is a waste of your time. The recovery phase represents a substantial deadweight loss of production to everyone, concentrated on the unemployed. And there are dynamic effects. Contact books get stale, and technology changes, so the longer people stay either unemployed or underemployed, the bigger the gap. This little model also gives us hysteresis.
But we can go further with the snakes-and-ladders model. Markets, we are often told, are information-processing mechanisms. Let’s look at this from a Diego Gambetta-inspired signalling perspective. The only genuinely reliable way to know if someone is any good at a job is to let them try. I have, after all, every reason to pad my CV, overstate my achievements, and conceal my failures. The only genuinely reliable way to know if someone is an acceptable boss is to work for them. They have every reason to talk in circles about pay and repress their authoritarian streak.
In a tight labour market, people move along close-to-optimal career paths. In doing so, they gain both experience, and also reputation, its outward sign. Importantly, they also gain information about themselves – you don’t know, after all, if you can do the job until you try. The same process is happening with firms and with individual entrepreneurs or managers. Because the information is the product of actual experience, it is costly and therefore trustworthy.
Now let’s blow the system up. We introduce a shock that causes a large number of basically random firms to fail and sack everyone. Because the failure of these firms is not informative about the individuals in them, the effect is to destroy the accumulated information in the labour market. Whatever the workers knew about Bust plc is now irrelevant. In so far as they’re now looking for jobs outside the industry, what Bust plc knew about them is also irrelevant. In the absence of information, the market for labour is now in an inefficient out-of-equilibrium state, where it will stay until the information is re-created. Walrasian tatonnement, right?
This explains an important point in Mason’s data that we’ve not got to yet. Why should people thrown off their career paths take much lower productivity jobs? You don’t need much information to know if someone can mow the lawn. In the post-crisis, disequilibrium state, low productivity jobs are privileged over high productivity jobs.
It strikes me that this little model explains a number of major economic problems. The UK’s productivity paradox, for example, is nicely explained by a huge compositional shift, in part driven by labour market reforms designed to make the unemployed take the first job-ish that comes along. Students who graduate into a recession lose out by about $100,000 over their lives. Verdoorn’s law, the strong empirical correlation between productivity and employment, also seems pretty obvious. Axel Leijonhufvud’s idea of the corridor of stability also fits. In the corridor, the market is self-adjusting, but once it gets outside its control limits, anything can happen.
And, you know, despite all the heterodoxy, it’s microfounded. Workers and employers are entirely rational. Money is just money. It’s not quite simple enough to have a single representative agent, because it needs at least two employers and two workers with dissimilar endowments, but it doesn’t need any actors who aren’t empirically observable.
It also has clear policy implications. If the unemployed sit it out and look for something better, you would expect a jobless recovery and then a productivity boom – like the US in the 1990s. If the unemployed take the first job-like position that comes along, you would expect a jobs miracle with terrible productivity growth, flat to falling wages, and a long period of foregone GDP growth. Like the UK in the 2010s. And if your labour market institutions are designed to prevent the information destruction in the first place, with a fallback to Keynesian reflation if that doesn’t cut it? Well, that sounds like Germany in the 2010s.
Call it Hayekian Keynesianism. Macroeconomic stabilisation is vital to keep the information-processing function of the labour market from breaking down.
That said, I wouldn’t be me if I didn’t point out that there are a whole lot of structural forces here that discriminate against specific groups. The post-crisis skew to low productivity jobs wouldn’t work, after all, if workers weren’t forced sellers of labour to capitalists. And there is one very large group of people who tend to get kicked off their optimal career path with lasting consequences. They’re about 50% of the population.