Unemploy = (
pdr.get_data_fred('UNRATE',
start = '1950-01-01')
.reset_index()
.rename(columns={'DATE': 'Date'})
.assign(UNRATE = lambda df: df['UNRATE']/100)
)Employment
Introduction
The Federal Reserve carries a dual mandate: 1) control inflation and 2) maintain full employment. In this section, we will focus on that second mandate and dive into the labor market.
Employment is perhaps the piece of the macroeconomy that influences people’s lives the most, or at least the most directly. The labor force in America is over 100 million people, and that doesn’t count people who used to be in the labor force and have retired (or people who are young and will someday be in the labor force). It’s also a piece of the macroeconomy that is quite tangible and understandable: we don’t need data science or economic theory to understand what happens when people start losing their jobs. Contrast that to interest rates (banned in ancient societies as evil!) or inflation (inflation is bad, but deflation is also bad?), which aren’t intuitive to most of us. Most people have a sense for how the labor market is trending, though we can be heavily biased by our own sectoral experience. If we work in data science, and the data science job market is doing poorly, we’ll naturally feel the entire labor market is doing poorly, even if other sectors like health care are booming. This stands in distinction to, say, interest rates or inflation or the stock market, which are experienced in a more fungible way.
Employment statistics are somewhat different from other common macroeconomic series in that they come from 2 separate monthly surveys: the unemployment rate comes from the household survey, while the payroll data comes from an establishment survey. Keep this in mind in the rare months when we have employment numbers that seem in conflict, such as a decrease in the number of payrolls but also a decrease in the unemployment rate.
With that in mind, we’ll make our way through employment data in the following way:
- Examine the unemployment rate
- Visualize unemployment and recessions
- Visualize payroll numbers
- Build the Sahm recession indicator
- Visualize job openings numbers
- Build a custom labor slack index
Unemployment Rate
Let’s start with the unemployment rate, which is defined by FRED as “the number of unemployed as a percentage of the labor force…restricted to people 16 years of age and older…who are not on active duty in the Armed Forces.”
Since this is an unemployment rate, the lower it gets the better news for the economy. Let’s pull the unemployment rate from FRED.
We chart the unemployment rate below along with our recession shading. Other than around the Covid shutdowns, the unemployment rate has been quite low from the Great Financial Crisis through 2023. In fact, we need to go back to the 1960’s to find another era with unemployment hovering around 3.5%.
(
alt.Chart(Unemploy,
title=alt.Title(
'Unemployment Rate',
subtitle = 'source: FRED'))
.mark_line()
.encode(
alt.X('Date:T')
.title(None),
alt.Y('UNRATE:Q')
.title(None)
.axis(format='%')
)
+
recession_shade
)
Note how recessions and unemployment rate spikes coincide and unemployment tends to peak after the recessionary periods have officially ended, and increases through the recession.
It’s easy to grasp the intuition of why unemployment is a lagging economic indicator. As the economic situation starts to deteriorate, a company may be reluctant to do layoffs because they are uncertain if we are truly in a recession or if it’s just a soft patch that will pass quickly. If it’s the latter and the firm reduces their headcount, they will have to rehire employees quickly, which is not easy to do and carries costs (retraining, hiring bonuses, etc.); hence, the company will wait until it is relatively convinced that the economy is in a recession before they engage in layoffs. When we are coming out of a recession we have the opposite case, where firms wait to hire more employees, and that is why the unemployment rate peaks after the recession has passed.
People tend to talk about unemployment in the context of recession frequently, so in the next section let’s build a table to help us communicate how those two relate to one another.