The Simple Math behind Immigration and Unemployment

Protectionists hate this one weird trick.

For every complex problem there is an answer that is clear, simple, and wrong.

-- H.L. Mencken

The Labour Force (LF) framework supposedly works like this:

1. If more people enter the labour force than get jobs, the unemployment rate goes up.
2. If less people enter the labour force than get jobs, the unemployment rate goes down.

Sounds fair, right?

I'll return to that in a second. 

It follows that if you starve the labour force of supply, say by cutting off immigration, the unemployment rate will go down, ceteris paribus (all other factors remaining the same). 

At least, one ABC opinion writer thinks so.

• Have closed borders pushed Australia’s unemployment rate down? This model shows they have (13-Feb-2022) 
• The unemployment rate is falling quickly. This 'special formula' explains why. (20-Feb-2022)

According to the LF framework, when we closed our borders it prevented people from coming to Australia to work and it saw thousands of migrants having to leave Australia, contributing to job vacancies jumping to record highs.

By closing the borders, it cut off a very large source of growth for the working-age population.

It meant the pace of growth in the working-age population fell dramatically below employment growth because the working-age population has only been able to grow from natural increase.

Employers haven't been able to source workers from overseas.

To be fair, the claim is only that the model shows working age population directly affecting unemployment. They provide a chart with which we're supposed to be able to make an eyeball confirmation.

But is that what actually happens? Historically, has it been as simple as 'more new employees than new workers means unemployment goes down'?

Not quite.

Worker Surplus vs ΔUnemployment

Between March 1978 and December 2022 there was a weak-to-moderate (0.4) correlation between surplus workers and change in unemployment, month-on-month.

A correlation coefficient ranges from -1 (perfect negative) through 0 (no correlation) to 1 (perfect positive). 0.4 is in the expected direction, but far too imperfect from which to draw conclusions. While rising worker surpluses did coincide with rising unemployment in many cases, there were many time periods when that did not happen. 

For example, December 1978, January 2005, and July 2020 were months during which more Australians were employed than reached working-age yet the unemployment rate increased.

No real-world relationship is perfect, naturally, but the one between unemployment rate and employment-to-working-age gap isn't even particularly good.

Again,

Surplus workers, perhaps counter-intuitively, translate little to unemployment.

So if the LF framework isn't as straightforward as is depicted, then that means migration - one of its many inputs - would have even less impact.

ΔNet-Overseas-Migration vs ΔUnemployment

Indeed, there is a weak negative (-0.29) correlation between change in Net-Overseas-Migration and Unemployment, year-on-year, between FY1977-1978 to FY2019-2020.

I would happily shout from the rooftops that more immigration means less unemployment, but I know that would be vague and inaccurate. For instance, I don't know if changes in immigration precede unemployment change or the other way around. Besides, the low magnitude of the correlation leads to the conclusion that:

Immigration, perhaps counter-intuitively, has little impact on unemployment. 

I don't want to cast the ABC as taking cheap shots at immigrants.

However, it is disingenuous for them to write,

Now, the argument I make in that video is not a comment on the benefits or otherwise of shutting the borders and cutting immigration in response to the pandemic.

, while in reality commenting on the logical impact of shutting the borders on employment.

That would not be controversial if we did not attach so much value to employment. We would not be up in arms about immigration's impact on, say, the price of paint. Linking immigration with employment, however baselessly, invites value-judgement on the otherwise neutral topic of immigration.

What's Going On?

First, the unemployment rate doesn't just measure the availability of workers. It also reflects behaviour. The out-of-work choose whether to seek jobs and employers choose whether to hire them. 

Second, the unemployment rate measures unemployment across the entire population, which is very large. While migration is a large part of new workers hitting the shelves, it is minuscule compared to the number of employed workers and the population in general. Population effects are overshadowed by greater trends, say, how well the workers on offer match the jobs on offer, and whether existing workers stay employed. 

With such a low correlation between worker numbers and unemployment, ceteris paribus is not a sound assumption and you really need to take a mutatis mutandis perspective. 

Then why is immigration continuously under the spotlight? You're probably expecting me to bring up racism. I actually think it's salience bias. It's easier to notice different-looking people moving in and out and between jobs, and conclude that they have an outsize impact on labour. Why? Because it's easier to notice different-looking people. Similarly large factors draw less attention if the people involved don't stand out. 

For instance, female employment growth in 2021 (+188,400) is roughly equivalent to net overseas migration in FY19-20 (+194,426). We do not speculate on the effect that the ingress of women into the labour force has on unemployment because female colleagues are no longer remarkable.

The truth is that everyone impacts unemployment, and there are big forces at play other than immigration.

So the simple math linking changes in working age population, employment, and unemployment is logically right but factually wrong.

Apparent reality is explicable not by math, but by bias.

How to use this one weird trick

The next time someone tries to use the Labour Force logic to 'prove' that increasing the workforce increases unemployment, tell them to run this correlation analysis across the data.

If they go further and insinuate that The Powers That Be are ignoring that inconvenient logic ...

According to some analysts, the closed borders have had some impact on the unemployment rate, but not a very big one.

RBA governor Philip Lowe made that very argument this month.

But I think that's wrong.

The closed borders have had a huge impact.

Remind them that sometimes things are dismissed by experts because they really are worth dismissing.

Notes

Here are the ABC datasets on Labour Force and Migration. I used seasonally adjusted figures when available.

Labour Force and Migration have different reporting periods: monthly and annually, respectively.

Why do I use change (Δ) instead of the base readings? 

1. It produces the strongest correlations. Believe it or not, I select against my inclinations.
2. I wanted to reduce the effects of long-term trends in population growth, declining unemployment, and increased migration.
3. I wanted to capture how the respective movements in each measure correlate.