"Moreover, the BLS states that the 90% confidence interval of the change in the number of jobs is approximately plus/minus 136,000." Wow that is a pretty big interval! I get it now :-)
I get your point about putting stats in context, and couldn't agree more. The challenge for people like me is that we don't know what contextual information is relevant, because we do not know the subject matter well enough or don't have the data. So for instance, with regards to the -0.5% change in GDP, I wouldn't know that it was driven by early buys ahead of tariffs. Hence very glad I read you! Great article, as always, thanks.
It's very true that headlines and articles skip the context. Especially in the context of labor markets, prior to reading some other economics substackers, I often saw the headline and trusted the conclusion (i.e. labor market is doing good/bad). Now I wait to see the write-up and analysis
What was left out of the article is the history she had for releasing flawed information for the previous administration which looks a great deal like partisan work. A month after she was fired, the BLS revised DOWN last years jobs report by 818,000 jobs. The report went from 93,000 to 911,000. Previous jobs reports seems to follow the same theme of being revised negatively as well. Partisan or incompetent we cant say.
"The BLS (Bureau of Labor Statistics) has released preliminary benchmark revisions for 2024, which show a downward adjustment of 818,000 jobs for the 12 months ending in March 2024. This is a significant, negative revision that indicates the job market was weaker than initially estimated. The final revisions, released in September 2025, revised the net job loss to 911,000 jobs for the period of April 2024 to March 2025. "
The information was not flawed. It's estimation. The BLS has a strict process for creating the estimates. As mentioned in the article, the BLS estimates a 160mln number. It's estimates change by around 0.1% to 0.2% of the total number of jobs at most.
The benchmark revision is a known process, which requires collecting significant data over the course of the year to make it a good benhcmark update. All of these steps are listed and known.
The 'narrative' issue is a what I consider a media issue - media, and many writers, misinterprets what the data can tell us, using catchy and meaningless headlines. For example, this statement "negative revision that indicates the job market was weaker than initially estimated" is a pure opinion with no basis in any economics research. We cannot say that because our estimate of jobs is lower, it means the labor market was weaker. You can easily have a stronger labor market with fewer jobs - the nature of jobs created, the growth in productivity, the change in population.
Thank you for the shout out! And nice post on the issues with these data. I'm thankful for people like you writing articles to educate others on how to think like an economist (and also how to think about data). We really need to push intro stats courses in high school to help people understand context and confidence intervals.
"Moreover, the BLS states that the 90% confidence interval of the change in the number of jobs is approximately plus/minus 136,000." Wow that is a pretty big interval! I get it now :-)
I get your point about putting stats in context, and couldn't agree more. The challenge for people like me is that we don't know what contextual information is relevant, because we do not know the subject matter well enough or don't have the data. So for instance, with regards to the -0.5% change in GDP, I wouldn't know that it was driven by early buys ahead of tariffs. Hence very glad I read you! Great article, as always, thanks.
Glad to hear you find my work helpful! :)
It's very true that headlines and articles skip the context. Especially in the context of labor markets, prior to reading some other economics substackers, I often saw the headline and trusted the conclusion (i.e. labor market is doing good/bad). Now I wait to see the write-up and analysis
What was left out of the article is the history she had for releasing flawed information for the previous administration which looks a great deal like partisan work. A month after she was fired, the BLS revised DOWN last years jobs report by 818,000 jobs. The report went from 93,000 to 911,000. Previous jobs reports seems to follow the same theme of being revised negatively as well. Partisan or incompetent we cant say.
"The BLS (Bureau of Labor Statistics) has released preliminary benchmark revisions for 2024, which show a downward adjustment of 818,000 jobs for the 12 months ending in March 2024. This is a significant, negative revision that indicates the job market was weaker than initially estimated. The final revisions, released in September 2025, revised the net job loss to 911,000 jobs for the period of April 2024 to March 2025. "
The information was not flawed. It's estimation. The BLS has a strict process for creating the estimates. As mentioned in the article, the BLS estimates a 160mln number. It's estimates change by around 0.1% to 0.2% of the total number of jobs at most.
The benchmark revision is a known process, which requires collecting significant data over the course of the year to make it a good benhcmark update. All of these steps are listed and known.
The 'narrative' issue is a what I consider a media issue - media, and many writers, misinterprets what the data can tell us, using catchy and meaningless headlines. For example, this statement "negative revision that indicates the job market was weaker than initially estimated" is a pure opinion with no basis in any economics research. We cannot say that because our estimate of jobs is lower, it means the labor market was weaker. You can easily have a stronger labor market with fewer jobs - the nature of jobs created, the growth in productivity, the change in population.
Thank you for suggesting my article!
Thank you for the shout out! And nice post on the issues with these data. I'm thankful for people like you writing articles to educate others on how to think like an economist (and also how to think about data). We really need to push intro stats courses in high school to help people understand context and confidence intervals.