Home > Uncategorized > Seasonal Adjustments Will Skew Labor Reports

Seasonal Adjustments Will Skew Labor Reports

This is a guest post by Aise O’Neil.

I think that the next two labor reports will overstate job growth because of a technical issue, namely Seasonal Adjustments. To explain why I will explain 1) what seasonal adjustments are and 2) why they will overestimate the strength of the labor market this year.

Explanation of Seasonal Adjustments:

The idea of seasonally adjusting data is that the meaning of a data point can depend on the time of year. For instance, say the government is reporting on a value, like fuel oil, which is known to go up in January each year as people use it more in the winter. If one is trying to notice trends in the price of fuel oil, they would like a data series on fuel oil that accounts for the usual January spikes and shows the unusual changes.

The mechanisms for calculating seasonal adjustments are complicated, vary by government department, and often require a few college courses on econometrics to understand. The essential idea is that governments can look at recent data to estimate how much higher or lower a value gets in a particular month relative to a long-term trend. For instance, oil prices might be 1%  higher in January than their longer-term trend according to recent data. When new seasonally adjusted data is reported, it will include the raw data plus an adjustment based on the estimates of how high or low the data is because of the time of year. For instance, when oil prices are released in the CPI report in January, the new seasonally adjusted data point may be 1% lower than the raw data in order to adjust for the fact that the raw data will show particularly high prices in January.

Seasonal adjustments to incoming data are made using estimates of seasonal trends which come from analyzing recent historical data. As more data comes in, seasonal adjustments to recent historical data can and will be retroactively revised.

For example, the Bureau of Labor Statistics (BLS), which releases both the CPI and Current Employment Statistics (“CES”) Report seasonally adjusts incoming data based on data from the past 5 years and revises it based on incoming data for the next 5 years. The CES is the source of the “unemployment rate” and monthly job growth figures which the media often reports on. The chart below is an example of BLS seasonal adjustments in action. It shows the seasonally adjusted and non seasonally adjusted estimates of the same thing: the population of employed people in the US. Without the seasonally adjusted data series we would normally see reports of employment going up or down based on the time of year, not underlying trends in the economy.

Why Seasonal Adjustments Will Overstate Job Growth

It’s pretty clear that seasonally adjusted data is going to give a better sense of what’s happening in the economy than data which isn’t seasonally adjusted. However, seasonal adjustments aren’t perfect. Seasonal adjustments to incoming data will be based on rigorous analysis of historical data and will implicitly presume incoming data will display the same seasonal trends as past data. I’d argue that the disruption Covid has made to schooling is going to disrupt seasonal trends in employment.

Normally the summer (and to a lesser extent winter) breaks will reduce overall employment for two reasons. Firstly, employment in education declines during summer and winter breaks. This can be seen in the graph below which depicts the seasonally adjusted and non-seasonally adjusted levels of employment in education. It includes data from 5 years prior to the Covid recession.

Secondly, parents have to look after their children and this will keep them out of the workforce. So employment levels should decline during school breaks. This would especially be true for women because women are more likely to be single parents and may bear more responsibility of looking after children generally even in two-parent heterosexual households. The graph below depicts the level of female employment in the United States divided by the overall level of employment. This is calculated using seasonally adjusted and non seasonally adjusted data series. It covers the 5 years prior to the Covid Recession. The relative female employment consistently falls in non seasonally adjusted terms going into the summer.

Both of these factors changed in the Covid era, because of virtual learning. In such an environment, certain seasonal jobs like janitorial staff, cafeteria workers, IT workers, and so on aren’t as prevalent as they used to be. The seasonal fluctuation coming directly from education is thus weaker. Additionally, many parents have had to stay out of the workforce to look after their kids throughout the year during online schooling. That means the seasonal fluctuation of summer break starting or ending is less important too.

With weaker seasonal effects now, seasonal adjustments which are based on historical data will over-account for them. When seasonal effects of the school holidays starting/ending will depress/increase employment, the seasonally adjusted levels of employment should show an increase/decrease.

One way to confirm this speculation is to look at what happens to both indicators of labor market health from August to September. That period is the strongest time for relative female employment growth as well as educational employment growth as a lot of schools start their fall semester in late August. As a consequence, the seasonally adjusted figures in both cases show a decline in seasonally adjusted terms. The graphs below show seasonally adjusted education employment and relative female employment over the past year to show the apparent weakness at the beginning of fall.

Both indicators of labor market health in seasonally adjusted terms fell significantly from August to September of 2020. This is because the start of the school year failed to create as many jobs for women and educators as it normally does, so in seasonally adjusted terms it showed a drop. In raw terms, the actual levels of relative female employment and employment in education increased, as is usual for that time of year.

In this same fashion, one can expect that the start of summer break will destroy less jobs than it normally does. As a result, in seasonally adjusted terms, the May and June jobs report will likely show strong job creation. This misleading seasonally adjusted data will be what the media reports in terms of the unemployment rate and monthly job creation. 

Categories: Uncategorized
  1. No comments yet.
  1. May 26, 2021 at 3:49 pm
  2. June 7, 2021 at 2:12 pm

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: