Metro Unemployment Rates (Or: When Summarized Statistics are Misleading)
So, the BLS just put out the new Metropolitan Area Employment and Unemployment Report (reporting for July). It looks good! About 90 percent of Metro areas saw an improvement in their unemployment rate.
So, I got curious. How is the Canton-Massillon area doing?
On the face of it, it looks good! The unemployment rate fell from 7.5% to 7.3% - a noticeable (if small) improvement.
But, then I look deeper. Why? Because of math and the way these statistics are calculated.
To get at that... We divide the civilian population into 3 categories: employed, unemployed, and not in the labor force. Employed people have a job. Unemployed do not have a job, but are actively seeking one. Not in the Labor Force neither have a job NOR are actively seeking one. So, my grandma who is not working and doesn't want to work is not unemployed. She is "not in the labor force".
Now, how is the typical unemployment rate calculated? Here we go!
Unemployment rate = Unemployed /(Employed + Unemployed)
So, the unemployment rate can drop in a number of ways:
(1) Unemployed people can find work. Here, the numerator falls while the denominator stays unchanged - the fraction falls.
(2) People not in the labor force can become employed. This is a weird case - as it requires people not looking for jobs to start looking AND find them quickly enough that they never show up as "unemployed" in a survey, but it can certainly happen. In this case, the numerator stays unchanged (same number are unemployed), but the denominator rises - the fraction falls.
(3) Unemployed people can leave the labor force. Here, both numerator and denominator fall by the same amount - so the fraction falls for provable mathy reasons. (Proof left to the reader.)
Both (1) and (2) would generally be considered "good" for workers - as they indicate that more people are becoming employed. But, (3) is not necessarily good - as it suggests that people can't find a job, so they're just giving up looking for work. This is probably not good.
What do things look like in Canton-Massillon? Well...
Labor force in June = 202.2K
Labor force in July = 201.8K
Unemployed in June = 15.2K
Unemployed in July = 14.7K
Employed in June = 187K
Employed in July = 187.1K
So, it's true that the number of unemployed fell by 500 people. But, 80% of that can be explained by the fact that 400 people left the labor force (or, possibly, moved out of the area). Only 100 found jobs.
To put it another way: let's say that those 400 kept looking for work rather than leaving the labor force. Based on employment changes alone, the unemployment rate would still have rounded to 7.5%, not 7.3%.
Ohio as a whole sees a similar - but not quite as bad - story. the number of unemployed in the state of Ohio fell by 11,700 - but 5,000 (just under 50%) of those can be explained from people leaving the labor force. (Some cities have very nice numbers - the nearby Cleveland-Elyria-Mentor had 14,400 JOIN the labor force, and unemployment fell by 3,700 people - suggesting 18,100 people found work who didn't have work before.)
IMPORTANT NOTE: There is a common, very wrong, belief that the "unemployment rate" is tied to the number of people claiming or collecting unemployment checks. So, I've heard some people say "Well, unemployment rates are falling because people hit the end of the X weeks that they can collect!" This is wrong. To be considered "unemployed" in BLS surveys, you must have these qualifications: (1) you have actively looked for work in the past 4 weeks. (or, alternatively, you expect to be called back from a temporary layoff) (2) you do not have a job. (3) you are currently available to work. That's it. You don't have to be collecting unemployment checks. More on the survey here, and on the relationship to unemployment claims here.
So, I got curious. How is the Canton-Massillon area doing?
On the face of it, it looks good! The unemployment rate fell from 7.5% to 7.3% - a noticeable (if small) improvement.
But, then I look deeper. Why? Because of math and the way these statistics are calculated.
To get at that... We divide the civilian population into 3 categories: employed, unemployed, and not in the labor force. Employed people have a job. Unemployed do not have a job, but are actively seeking one. Not in the Labor Force neither have a job NOR are actively seeking one. So, my grandma who is not working and doesn't want to work is not unemployed. She is "not in the labor force".
Now, how is the typical unemployment rate calculated? Here we go!
Unemployment rate = Unemployed /(Employed + Unemployed)
So, the unemployment rate can drop in a number of ways:
(1) Unemployed people can find work. Here, the numerator falls while the denominator stays unchanged - the fraction falls.
(2) People not in the labor force can become employed. This is a weird case - as it requires people not looking for jobs to start looking AND find them quickly enough that they never show up as "unemployed" in a survey, but it can certainly happen. In this case, the numerator stays unchanged (same number are unemployed), but the denominator rises - the fraction falls.
(3) Unemployed people can leave the labor force. Here, both numerator and denominator fall by the same amount - so the fraction falls for provable mathy reasons. (Proof left to the reader.)
Both (1) and (2) would generally be considered "good" for workers - as they indicate that more people are becoming employed. But, (3) is not necessarily good - as it suggests that people can't find a job, so they're just giving up looking for work. This is probably not good.
What do things look like in Canton-Massillon? Well...
Labor force in June = 202.2K
Labor force in July = 201.8K
Unemployed in June = 15.2K
Unemployed in July = 14.7K
Employed in June = 187K
Employed in July = 187.1K
So, it's true that the number of unemployed fell by 500 people. But, 80% of that can be explained by the fact that 400 people left the labor force (or, possibly, moved out of the area). Only 100 found jobs.
To put it another way: let's say that those 400 kept looking for work rather than leaving the labor force. Based on employment changes alone, the unemployment rate would still have rounded to 7.5%, not 7.3%.
Ohio as a whole sees a similar - but not quite as bad - story. the number of unemployed in the state of Ohio fell by 11,700 - but 5,000 (just under 50%) of those can be explained from people leaving the labor force. (Some cities have very nice numbers - the nearby Cleveland-Elyria-Mentor had 14,400 JOIN the labor force, and unemployment fell by 3,700 people - suggesting 18,100 people found work who didn't have work before.)
IMPORTANT NOTE: There is a common, very wrong, belief that the "unemployment rate" is tied to the number of people claiming or collecting unemployment checks. So, I've heard some people say "Well, unemployment rates are falling because people hit the end of the X weeks that they can collect!" This is wrong. To be considered "unemployed" in BLS surveys, you must have these qualifications: (1) you have actively looked for work in the past 4 weeks. (or, alternatively, you expect to be called back from a temporary layoff) (2) you do not have a job. (3) you are currently available to work. That's it. You don't have to be collecting unemployment checks. More on the survey here, and on the relationship to unemployment claims here.