So Franklin and Delaware Counties are bright spots in a state where most counties lose population overall, but what other information can we infer from these growth rates? A friend asked if these growth trends might indicate that individuals in Central Ohio are starting to prefer urban environments over the suburban, with Franklin County representing urbanity and Delaware County the suburbs. It’s a good question, but the short answer is… not really.

Beyond the fact that one year is not enough information to establish any sort of statistically relevant trend, that more people are moving into Franklin County matters a little less than where in Franklin County they are moving (as some have already discussed). After all, there’s a big difference between greenfield development outside of Columbus City limits and moving into the city center. It’s also important to separate natural population changes from migration. Natural population growth includes births and deaths, and the remainder are individuals moving in and out of an area.  This type of voluntary population change would obviously be the more interesting demographic for exploring a preference for urban or suburban environments.

But getting back to my friend’s question, there are some basic ways we can look at this data to get a rough idea of where things stand. Again the really interesting question isn’t whether people are moving into Franklin County, it’s where they’re going. We can get an idea of exactly where people are going in the county by disaggregating the Census data from the entire county to census block groups. Since there’s no question on the US Census about a preference for urban or suburban communities, the next step is picking some kind of proxy. I decided to use population density. Population density is usually expressed by the number of people per square mile or square kilometer. Basically, it is the ratio of people to space: a small area with a large number of individuals has a higher population density than a large area with few individuals. Higher population densities, therefore, correlate with more urban environments and lower population densities with suburban or rural areas. My thought was that if people in central Ohio were starting to prefer urban communities, then the population density of Columbus proper should increase over time. Sounds like a perfect excuse to make a few maps.

Before I get to the maps, I have to emphasize that this is a quick survey exercise. I’m an academic at heart and I’d feel horrible if I didn’t point out that I didn’t triple check my numbers, I didn’t account for natural vs. migratory population changes, and I didn’t account for the growth rate (percentage increase) of each census block, or any number or time consuming things that would better validate these results. This is a sketch of population patterns in Franklin County, Ohio, and it’s a game anyone can play. If you have some kind of access to a mapping program, I encourage you to download this free data from American FactFinder and explore on your own. It’s a fun nerdy time.

First we have Franklin County’s population density in the year 2000.  I calculated population density by dividing the population of each census block group by its area.  If I were making this map again I would probably leave the numbers in the population density scale, but at the time I thought the numbers were too confusing because the census block groups can be such small areas.  I’d also retain labels for the highways and major roads to make the map easier to understand.  Regardless, I think this map largely reflects what you’d expect if you’re familiar with Columbus or Franklin County.  The areas with the highest population density are generally within I-270 (the circular outerbelt highway) and are further concentrated immediately south and north of downtown (downtown is smack in the middle of the map within the rectangle of highways).

Next is Franklin County’s population density in 2012 using the census estimates.  You can’t see it on the map, but I decided to keep the numeric range behind the density scale (lowest, low, middle, high, highest) the exact same as in the 2000 map because I wanted to see exactly how the population densities did or did not change.  My theory was that, given the same scale, an influx of people into the city (a preference for urban living) would result in more dark blue areas around the core.

It might be a bit difficult to see at first, but instead of a steady gain in population density in the heart of the city, it seems that the population densities spread out a bit throughout the county, meaning there was actually some density loss in Columbus, particularly in the German Village/South Side area.  (It would be interesting to go through foreclosure data to see if this area was particularly affected by the housing crash – perhaps that can be a future map series.)  North of downtown in the Short North/University area held pretty steady and density gains are apparent in the communities that surround Columbus proper.

After making these maps to demonstrate why an increase in Franklin County’s population doesn’t necessarily mean an increased desire for urban living, I started to wonder if I was over complicating the issue.  Maybe it would be better to simply see how areas lost or gained population over the twelve years.  I decided to make another map that just looked at whether each census block group lost population, gained population, or held steady from 2000 – 2012.  Again, I did not differentiate between natural and migratory population change.

The green areas in this map represent a net population gain since the year 2000 while the grey displays a net population loss. This map largely confirms the 2012 population density map in that the areas that lost density experienced a net loss of population over the same time period.  It also suggests in simple binary terms that areas outside of Columbus seem to have experienced population growth in equal or greater terms as the city proper.  To double check I made one more map highlighting Columbus’s city limits.

The opaque areas of the map are Columbus proper.  This makes it a bit easier to see that although the city itself has experienced a healthy amount of growth over its entire geographic area, there were losses in eastern portion of the city and much of the growth is in the periphery.

At the end of the day, these maps confirmed my suspicions: Sorry, friend.  Franklin County’s growth rate doesn’t really mean that more people are choosing an urban life style.  However, the last map brought a bit of unexpected optimism.  Columbus may not be the densest urban environment, it may have even lost some population density since 2000, but it has experienced positive growth downtown and downtown’s surrounding neighborhoods.  I can’t help but feel that this is a good sign; if you’ve seen the development boom in central Columbus lately, you might agree.  From the mixed-use development underway at High and Hubbard in the Short North, to the construction of apartments around Columbus Commons, to the the grand opening of the downtown Hills Market, it certainly feels like there’s a renewed momentum for the central area of the city.    Perhaps in another ten or so years we’ll look at the census data and see, thanks to present day efforts, that individuals are in fact expressing a preference for urban living in Central Ohio.

I’ve always been fascinated by tornadoes.  They take on many different shapes and sizes and can be quite beautiful.  But tornadoes are serious business.  Researchers and chasers study them relentlessly.  They have their own reality television shows.  The art and science of predicting where and when a tornado will strike has improved greatly since 1974, but there is still much we don’t know about tornadoes.

I’m sure at one time you’ve seen a traditional “Tornado Alley” map or maybe you’ve seen a map of the U.S. counties most likely to get hit with a tornado.   I wanted to create a map that was more detailed than something at the county level.  I wanted to zero in on precise locations where tornadoes have historically occurred because the past is likely to predict the future.

To start, I located some data provided by the National Weather Service (NWS).  They had a GIS file of tornado tracks from 1950-2006.   Information on the intensity (EF scale), the length and width of the track, property and crop loss estimates, as well as fatalities and injuries were included in the file’s attributes.  In order to quantify the impact of a tornado without including biased data,  I chose two variables  – the number of tornadoes and the intensity of each tornado.  Next, I simply laid out an imaginary 10 square mile grid across the U.S. as a geography for aggregating my data.  I chose a 10 square mile grid because it is usually much smaller than a county (on average you can fit 4-5 grid cells within an average sized county).  I counted each tornado that crossed into a grid cell and summed up the EF scale intensity of each tornado (actually, I added a value of 1 to each storm’s EF number to account for storms with an intensity of EF 0 ).  Each of the data values were normalized before computing a final value for each between 0 and 1.

The results of the exercise can be found here in this interactive map.  Based on our methodology, the part of the country most likely to experience a tornado is located on the Oklahoma and Kansas border – specifically, the the northwest corner of Kay County, OK and the southeast corner of Sumner County, KS:

Luckily, this is not a densely populated area.  In fact, less than 500 people live in this particular cell.  However, the Top Ten Tornado Hot Spots include several areas where the population is high:

As you can see, there are several heavily populated corridors that are historically most likely to experience a tornado.  Oklahoma City (OK), Shreveport (LA), Little Rock (AR), and Jackson (MS) are the most heavily populated cities within our computed danger zone.

If we assume that small changes in the climate over time will not result in dramatic shifts of tornadic activity, then we can safely predict that the areas of high tornadic activity in the past will continue to experience intense, long-track tornadoes into the future.  This knowledge should affect things like building design and cityurban design, disaster preparedness, and insurance rates.

We’ll be posting various maps related to this exercise on our Pinterest site over the next couple of weeks.  Check back from time to time to see what we’ve come up with.

I don’t think there is any powerful basketball inference you can make regarding the location of these schools.  However, nobody is more concerned about the location of these schools than CBS Sports because this could be a ratings black hole.  Baylor (Waco, TX) and Kansas (Lawrence, KS) are the westernmost schools in the Sweet Sixteen.  Syracuse is the closest school to the largest media market in the U.S. – New York City.

History has shown that the higher seeds bring in higher ratings.  Therefore, we can assume that CBS is rooting against the likes of Ohio University, NC State, Xavier and UC.  So while it may be exciting for us Ohioans to have four teams represented in this year’s Sweet Sixteen, CBS wants the madness to end no later than Friday evening.  One thing is for sure, there will be at least one less Ohio team after the next round – Ohio State plays Cincinnati in the East Region Semifinal in Boston on Thursday night.

I’m not advocating for a four-year degree for everyone, but the data is clear and the facts are unavoidable – you are in a substantially better position professionally (and economically) if you have at minimum a Bachelor’s degree.  Of course, this club has some obvious barriers to entry.  The two most obvious are cost and aptitude.  But another potential barrier is location – how far must one travel to attend an institute of higher learning?  Bill told me about an initiative that former Ohio governor James Rhodes had championed several decades ago.  Governor Rhodes wanted every Ohioan to live within 20 miles of a college or university.  That gave me an idea.  I wanted to see where this at-risk population lived in relation to the location of colleges and universities – hence this installment of Urban Decision Group’s Map of the Week series.

Colleges and universities were defined as anything having a NAICS code of 61131009.  The data was extracted from a business database provided by InfoGroup.  I don’t assume 100% accuracy with any third-party data sets, but the data we use from InfoGroup is actually pretty good stuff.  They provided point data geocoded to the address of the institution.  I then established 10-mile rings around each point.  Normally, if I were establishing a trade area, I would never use a simple ring around a point.  But we can get away with it in this case because of the shear volume of points create several areas of overlap.  The 10-mile radius around each college and university represent  areas that we are not concerned about.  The areas we are interested in are everything outside of these rings; they represent population centers that are more than ten miles away from an institution of higher learning.  So I laid out a 10 square mile grid across the U.S.only for those areas that were not within 10 miles of a college or university.  This area represents territory where location could prove to be a barrier to higher education.

The next step was to define what the undereducated at-risk population actually is.  The data was extracted from the American Community Survey (ACS) 2006-2010 data at the county level and ultimately aggregated into the 10 square mile grid cells.  I decided to focus on the age group of 35-64.  People in this age group are generally less mobile than young people.  This group consists of households with children, mortgages, and many other things that prohibit a semi-transient lifestyle.  Then I broke the data into three sets.  The first set consists of those people without a high school diploma.  The second set contains those with no college and just a high school diploma.  The final set was simply the percent of the population that only had a high school diploma.  The logic in choosing this data is that no single data set could define what the at-risk population was, but the combination of the  three would provide a pretty good definition.  Each of the data sets was normalized and a final normal score was calculated for each grid cell.  Normal score values are guaranteed to fall between 0 and 1.  A value trending towards 1 indicates more of the population is at-risk.

When viewed on a map, we can identify the location of the undereducated at-risk population.  If members of this population group were to become unemployed, they are the most at-risk for prolonged periods of unemployment.  You can make the argument that with the ubiquity of the Internet and the rise in online courses available through many colleges and universities, location no longer matters.  This may be true for a small subset of the population but the at-risk population that we identified is less likely to have high-speed internet or even awareness that such opportunities may exist.

Like Urban Decision Group’s previous Maps of the Week, our intent is not only to inform but to inspire.  Decision and policy makers can direct resources more efficiently if they have a clear picture illustrating where they should go.  This week’s map is no exception.  Again, I’m not advocating that everyone in this population group needs a four year degree.  But at minimum everyone should have reasonable access to technical job training and vocational schools.  Education not only benefits those that receive it, but improves the health of the entire economy.  The proof is in the gap between unemployment rates for the educated and the undereducated.

The map uses county data from the 2006-2010 American Community Survey (ACS) and is ultimately aggregated into 10 square mile grid cells.  There were two criteria used in calculating a popular location “score”.  First, we looked at the total number of workers that work from home (telecommuters) in each U.S. county.  Then the data was normalized.  Normalization is the process of ranking the data on a scale of 0 to 1 using the county with the most telecommuters as the base.  The top county gets a score of 1 and all other counties are scored in proportion to the top county.  For example, Los Angeles County, California had 200,450 people working from home; therefore, they received a score of 1 for this category.  Maricopa County, Arizona was second on the list with 88,689 people working from home.  Their normalized score is 0.44 which was calculated by dividing the number of commuters in Marcopa County (88,689) by the top value from Los Angeles County (200,450).  This step was repeated for each county to produce a normalized telecommuting score.

The top ten counties in terms of total number of people working from home are:

1. Los Angeles County, CA – 200,450
2. Maricopa County, AZ – 88,689
3. Cook County, IL – 88,287
4. San Diego County, CA – 86,297
5. Orange County, CA – 66,404
6. Harris County, TX – 57,861
7. King County, WA – 53,621
8. New York County, NY – 52,281
9. Riverside County, CA – 41,753
10. Miami-Dade County, FL – 41,560

The second category we looked at was the number of people working from home as a percentage of all workers in the county.  Analyzing the data in this fashion allows us to pay proper attention to those counties that are not as heavily populated, but yet have a high percentage of workers telecommuting.  The top county in this category is Wheeler County, Nebraska which had 40.45% of their workers working from home.  This data was also normalized.

The counties with the highest percentage of the workforce working from home are:

1. Wheeler County, NE – 40.45%
2. Chattahoochee County, GA – 39.24%
3. Slope County, ND – 38.19%
4. Arthur County, NE – 32.88%
5. Pulaski County, MO – 32.54%
6. Billings County, ND – 30.51%
7. Kidder County, ND – 29.20%
8. Carter County, MT – 28.83%
9. Harding County, SD – 28.11%
10. Loup County, NE – 27.76%

The final score used in our map is  simply the combination of these two scores for each county divided by two.   This allows us to give equal weight to both data categories.  The final top ten counties are thus:

1. Los Angeles County, CA  (normal score = 0.56)
2. Wheeler County, NE (normal score = 0.50)
3. Chatahoochee County, GA (normal score = 0.49)
4. Slope County, ND (normal score = 0.47)
5. Pulaski County, MO (normal score = 0.42)
6. Arthur County, NE (normal score = 0.41)
7. Billings County, ND (normal score = 0.38)
8. Kidder County, ND (normal score = 0.36)
9. Carter County, MT (normal score = 0.36)
10. Harding County, SD (normal score = 0.35)

The final step was to apportion the data into 10 square mile grid cells.  This final step accomplishes a couple of things.  First, it makes it quick and easy to display on a web map.  Second, it ignores political boundaries by considering  data from surrounding counties.  The result is a thematic map that displays the most popular locations for telecommuters.

Urban Decision Group (UDG) is responsible for the creation of this map.

Last week we calculated the average commuting time in minutes per trip.  From this we can calculate the the average number of hours a worker spends commuting per week regardless of the mode of transportation.  We then used two ACS datasets to compute the average hourly wages for a worker that lives within each grid cell (a 20 square mile area).  The  average hourly wage computation is based on a 40 hour work week and 48 weeks of work per year-this assumption factors in holidays and 2 weeks of vacation per year.  We then computed a new hourly wage by adding the number of hours spent commuting per week to 40.  In other words, if the average worker spends 3 hours per week commuting, then they are actually “working” 43 hours per week.  The base hourly wage is then subtracted from the “new” hourly wage (which factors in time spent commuting) to arrive at the decrease in hourly wages after factoring in the time spent commuting.

If you click on any cell in this week’s map, it will display the amount of dollars you can subtract from the average worker’s hourly wage to derive the actual hourly wage after considering the average worker’s commute time.  Why is this important?  It’s important because it considers things like negative externalities and opportunity costs.  The time spent commuting to a job is time that the worker could be spending with friends or family, working on another job, etc.  In other words, a worker may want to consider a lower paying job if the job is closer to hisher home.  This particular exercise is exclusive of gas prices, which is a whole other matter.

Keep in mind, the grid cells represent the location of the worker’s residence, not the location of employment.  The location of employment has already been factored in when we computed the time spent commuting.  When you click on a cell, the field named CHGINWAGES contains the dollar amount you should subtract from the worker’s hourly wage.  If you want to compute your actual hourly wage you first need to compute your base hourly wage by taking your salary, divide it by 40, and then divide that by 48.  Then locate the grid cell that represents your residence, click on it, and then subtract the amount displayed from your hourly wage.