There's no shortage of information about cities. From population and unemployment to crime and traffic statistics, urban and regional data is swirling all around us. We're exposed to ranking after ranking, from the best places to raise families, to the metro areas with the best public transportation or local restaurant scene, all the way down to less tangible measures like say, the worst-dressed cities.
But do we really internalize all this information in a meaningful way? Are our opinions of cities actually affected by statistics and rankings? Or is popular sentiment toward a city based more on nebulous, harder-to-quantify properties? Furthermore, how consistent are our opinions of cities? Do they vary from region to region?
Along with Andrew Mauboussin, a research assistant at the Kauffman Foundation and a rising freshman at Harvard University, I set out to add a bit of rigor to these questions by creating a simple city sentiment survey. I asked 310 individuals from the United States for their mental reactions when they hear the names of the 50 largest metropolitan regions in the U.S., and recorded their home ZIP codes.
Before we get to the results, some notes on methodology: I found my survey respondents via Amazon Mechanical Turk, an online labor market designed to allow the completion of tasks that are hard for computers to do but easy for people. Have a lot of photos that need captions? There are respondents on hand, called Turkers, who will complete these tasks for very small amounts of money. Recently, social scientists have realized that MTurk can be an amazing testbed for social experiments, ranging from behavioral economics to network science. It can also be used for surveys.
That said, MTurk is by no means a representative sample, but it can at least give us a starting point for thinking about our questions. So on a scale from one to nine (very negative to very positive), I asked 310 Turkers how favorably they viewed the country's largest metro areas. We then gathered up their responses and looked for patterns.
The overall rankings by all 310 respondents are mapped above. Blue indicates a more positive opinion, while red is more negative. The size of the circles are scaled according the actual ratings.
What we found is that our initial perceptions about cities are in fact often grounded in statistical reality. The positive or negative opinions of our survey respondents were correlated, often quite strongly, with such metrics as change in population, housing prices, and cost of living, and inversely correlated with measures like crime and unemployment. On the other hand, measures such as sales tax and traffic congestion appear to have little influence on people’s perceptions of different cities.
So, which cities have the most positive associations? Overall, the top ten cities among our survey respondents were the following:
10. Austin, Tex.
9. Charlotte, N.C.
8. Orlando, Fla.
7. Boston, Mass.
6. New York, N.Y.
5. Portland, Ore.
4. San Diego, Calif.
3. San Francisco, Calif.
2. Denver, Colo.
1. Seattle, Wash.
And the ten cities with the highest percentage of negative reactions:
10. Salt Lake City, Utah
9. Milwaukee, Wis.
8. Columbus, Ohio
7. Jacksonville, Fla.
6. St. Louis, Mo.
5. Cincinnati, Ohio
4. Oklahoma City, Okla.
3. Cleveland, Ohio
2. Birmingham, Ala.
1. Detroit, Mich.
Overall, Detroit rates far lower than any other city, making it difficult to do certain analyses because it actually drags everything else down. However, certain patterns jump out.
If you look at the different regions of the United States, as divided by the Census Bureau into Northeast, Midwest, South, and West, we can see how respondents from different regions rank the top 50 American cities.
Every region rates coastal cities high, with Denver the only inland city making the top five according to any region. On the flip side, all regions rate inland cities near the bottom, with only the Midwest including two cities from the coast – Houston and Los Angeles – in its bottom five.
Further, if we search for cities that yield significantly divergent opinions depending on the region that's rating it, certain cities jump out. Just for example, San Francisco shows distinct sentiment differences depending on the region doing the rating. Respondents in the South and Midwest have less favorable views of the Bay Area city than it enjoys in the Northeast and West.
While certain cities are positively viewed by all regions, each region has a better view of its own cities than those cities of other regions. The South likes southern cities, the West western cities, and so forth. The Midwest appears to be the most self-hating (or at least the least positive toward itself) of the Census regions.
Of course, there are many more questions that can be asked of our respondents. Why do you like or dislike this city? Do you feel that your opinions are based on fact or something less well defined? But even without knowing any of these things, it seems that on the whole we're fairly adept at integrating the multitude of the city-based information we're exposed to on a regular basis into a snap judgment—though we Midwesterners have our work cut out when it comes to urban self-esteem.