Unemployment Across the States

So we already know that unemployment is high. We're becoming startlingly aware of how high long-term and youth unemployment is as well. A next step, especially when we're trying to see what policies helped unemployment most, is to visualize how state unemployment rates differ. Has the recovery picked up in some places, but left others behind?

Over on FiveThirtyEight, Ben Casselman has an answer to that question. Unfortunately his conclusion is it "doesn't tell us much." He looked at a couple of ways of thinking about the spread of jobless rates across states, both of which showed a similar trend, and settled on a ratio of the highest and lowest state rates. Outside of an oil price shock in the '80s, the ratio is relatively steady since the '70s. 

I understand why he did it this way - it's a relatively straight-forward way of visualizing how states differ. After thinking about it, and especially about low unemployment states, that ratio loses a lot of valuable information. By including the lowest unemployment rate (which has averaged 2.8% since 1985), it's including something not that interesting. Today's lowest unemployment is in North Dakota - which has less than 800,000 residents and an oil boom. That's tiny, and most of  the country does not have an oil boom. Most of the other states with the lowest jobless rate have been similarly uninteresting for the rest of the country (mostly energy or commodity booms in states with small populations).

This presents a tough question for telling stories with data: how much can you see the forest for the trees without losing something interesting? As the first real data-driven news website, FiveThirtyEight is grappling with this trade-off more than most. For economic issues, they've done a pretty good job of it.

There are still 48 states and the District of Columbia that get skipped in the story. Certainly there's something interesting in those. To see if there was, I set up a box-and-whisker chart. It was my favorite one in middle school, and it tells a lot about the spread of jobless rates that a single line can't. The green bars cover the rate of the middle 50% of states. The whiskers extend all the way to the most extreme value.

Using BLS data, I constructed the chart for statewide unemployment rates since 1985 (I wanted to avoid the oil price spike in Casselman's graph, so I could see if there's anything interesting in the intervening years). I averaged each state's monthly rates to get a yearly average. 

The spread between the middle 50% of states was shrinking, until the Great Recession hit.

The spread between the middle 50% of states was shrinking, until the Great Recession hit.

What's most clear in this chart is how the spread became larger after the Great Recession. The middle 50% differed by 1.4% in the five years leading up to the recession, then 2.6% after. The recession did not hit all states evenly. It didn't even just hit the most average states evenly. The bottom half of states separated from the top half. You can see this in how the median moves lower in the middle 50% since 2008. It used to be pretty much in the middle, meaning that the spread was fairly even. The 12th worst state and the 12th best state were the same % away from the median. But the 12th worst state became relatively worse since 2008.


By now, we all know that the recession has been worse on some parts of the population - long-term unemployed, those without college degrees, youth - but it has also had a bigger effect on some states as well. It hasn't just been on the margin either. It's not just that Rhode Island (with a nationwide high of 8.9% unemployment) that is doing worse and North Dakota that is excelling. The just-worse-than-average have gotten worse in the meantime, places like Tennessee and South Carolina.

As the recovery has plodded on, this gap has eroded. Now the spread looks more like the pre-recession level, at least in terms of the worse-than-average getting better a little faster than the better-than-average. But the overall spread is still larger than we've seen since the mid-'80s.