Sunday, October 15, 2006
Fun with salary statistics
Yes, I just used 'statistics' and 'fun' in the same sentence. Don't let that stop you from reading on.
An article came out recently that talked about using U.S. Bureau of Labor Statistics information to compare salaries across location. According to the article, nurses get paid the most in San Jose, CA. Correctional officers make the most in Vineland, NJ.
"Well, sure they get paid a lot in San Jose--it costs a lot to live there," you might be saying. (You might also be saying, "What are you doing blogging on a Sunday?" or "Go Tigers!") Yes, this is true, and it's what I thought too. But surely there must be a way to compare jobs across the country taking cost of living into account...So I proceeded to attempt to find a tool to do just that...and failed.
So I started thinking...
What would you need to do to figure this out? Well, first you'd need salary information by occupation by geographic region. Fine, the BLS has data we can use to do that (click on "Metropolitan Area Cross-Industry Estimates"). And because we're smart cookies , we'll use median annual salary rather than mean because medians give a better picture of the "typical" number.
Next, we need cost of living data. You can get this from several places, including Yahoo, but I prefer Sperling's Best Places which offers all kinds of nifty information broken out by city.
Now all we need to do it match them up. Here's where our first challenge comes in. Cost of living data is broken down by city, salary data by metropolitan area. So what I did was average (and here I used the mean) the cities that make up that metropolitan area. No, it's not perfect, but it'll do.
Second challenge: how do we use these numbers? There are lots of ways, but as I always say, when in doubt, divide something by something else. In this case, I divided median annual salary by cost of living. This gives us some sense of the "true" salary when compared to where the job is. The higher the cost of living, the lower the number.
So what happens when we do this? Well, first of all, we wonder why someone hasn't done this before. But then, we do a sample job to see how this all works out. For my example I chose (purely at random) "Human Resources, Training, and Labor Relation Specialists." Let's start with how BLS displays the salary data. When we look at the BLS results, and these are available for all occupational groups , we see that salary is highest in these locations:
1. Framingham, MA ($80,070 mean annual wage)
2. Bridgeport-Stamford-Norwalk, CT ($76,200)
3. San Francisco-San Mateo-Redwood City, CA ($71,900)
4. Durham, NC ($69,260)
5. Wilmington, DE ($68, 210)
What happens if we use medians? Pretty much the same results, except the #5 spot goes to Hartford, CT and Wilmington slips to #7. Must be some HR specialists in Wilmington making bank.
Finally, what happens when we adjust median salary for cost of living? Let's call that the Bryan Index. Here are our top contenders (which I limited to approximately the 30 top salary cities--this stuff is time consuming, after all):
1. Niles-Benton Harbor, MI (median salary $63,150, Bryan Index of 842)
2. Durham, NC ($69,260 and 700)
3. Kennewick-Richland-Pasco, WA ($60,390 and 647)
4. Longview, WA ($62,750 and 644)
5. Wilmington, DE ($68,210 and 640)
So Durham and Wilmington hang in there--they have a good combination of high salary and relatively low cost of living. But the list has three new contenders, who are there because their cost of living is so competitive. Framingham drops to #10 due to its high cost of living.
This type of information is very useful, in many ways. Aside from being a great source of salary survey data, they help us answer this question: Are folks really underpaid or is that just the perception? Not to mention: Wow, we pay pretty well all things considered--are we doing anything with that information, like, I dunno, advertising it?
This is usually where someone says, "Hey Bryan, there is something that does this. Just go to www.youdidntgooglehardenough.com" Oh well. I had fun with statistics.