Opinion | How We Calculated Where to Live – The New York Times
Places
We started with a list of about 30,000 nonoverlapping places, as defined by the 2015-19 American Community Survey, which is conducted by the U.S. Census Bureau. (From here onward, any data we describe as coming from the census refers to that survey.)
People’s concept of place varies by state. In the Northeast, many people live in small villages or hamlets, which belong to larger townships that handle governmental functions. Sometimes, the census refers to those townships as “county subdivisions,” not “places.” After consulting with the Census Bureau, we decided to use a mix of places and county subdivisions for Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island and Vermont, based on the populations of places and the subdivisions containing them.
We excluded places with fewer than a thousand people, which had large margins of error for many statistics. We wanted to include places in Puerto Rico and other U.S. territories, but most of our data sets didn’t extend to those areas.
Because the goal of the quiz is to help people find places where they could move, we also filtered out places where a large share of the population lives in group quarters. This eliminated a handful of places where most residents are incarcerated, students or members of the military.
After all that, we ended up with a list of almost 17,000 places. They contain 77 percent of the U.S. population.
After clipping the boundaries of each place to exclude water features, we calculated population-weighted centroids for each place using WorldPop population data.
The drive time isochrones we used in our analyses were created using Here. They do not account for traffic.
Affordability
The affordability ratings ($-$$$$) are based on the mean percentile rank of three metrics: median household income, according to the census; median housing price, based on listings from Realtor.com over the past two years; and the average price per square foot of those housing listings. For the housing metrics, we included data only for places with 30 or more listings.
Each “$” bin has roughly the same number of people in it. (More than 7,000 places fall into the “$” bin, but most of those places are very small.)
The affordability ratings do not directly measure cost of living or taxes.
Size
The big cities filter excludes places with fewer than 250,000 people.
Race and ethnicity
We included scores for race and ethnicity to help readers who might otherwise struggle to find places with people who may be more likely to share their experiences and interests.
The scores are based on census categories, which are broad. A place with many Asian people could be home to second- and third-generation Chinese Americans, or it might be an enclave for immigrants from Pakistan.
The scores account for both the percentage and the number of people of each race in a place. Say a reader selects the “Black” checkbox. A town of 1,000 people where 60 percent of residents are non-Hispanic Black — far higher than the 2019 national average of 13 percent — will score highly. So will a large city with hundreds of thousands of Black people, even if its percentage of residents who are Black is the same as the national average.
The Hispanic score is based on the share and number of people identifying as Hispanic or Latino of any race. The Black, Asian, Native American and Pacific Islander scores are based on the share and number of non-Hispanic people of those races.
The racial diversity score is based on a place’s multigroup entropy index, which measures how balanced a place is in terms of different races and ethnicities. A place with an equal number of white, Black, Hispanic, Asian, Pacific Islander and Native American people would get a perfect score. The score also factors in the share of residents who are not non-Hispanic white people.
We’ve normalized the racial diversity scores so that places where a few races constitute a significant fraction of the population get full credit. The score does not measure how spatially integrated people of different races are within a place.
Age
Like the race metrics, the scores for young adults, parents and retirees account for both the percentage and the number of people in a place, according to the census.
The young adults score measures the share and number of people ages 20 to 34.
The parents score is based on the number of households that include someone under the age of 18.
The retirees score is based on the share and number of people 65 years old or older who are not in the work force.
Climate
The scores for temperature, snowfall and sunniness are based on 20-year averages from the National Oceanic and Atmospheric Administration weather station closest to the population-weighted center of each place, according to AccuWeather. Some rural places will have less precise scores because the nearest weather station with 20 years of data is many miles away. We decided that imprecise data was still useful for those places, even if our values may differ slightly from actual conditions.
The warm winter score is based on AccuWeather’s lowest average monthly low RealFeel temperature during December, January and February of each year. The RealFeel combines temperature, humidity, cloud cover, sun intensity and wind to measure how hot or cold a place feels.
The cool summer score is based on the highest average monthly high RealFeel temperature during June, July and August of each year.
The snowfall scores are based on the total accumulated snowfall during December, January and February of each year.
The sunniness score measures the share of minutes during the day during which the sun was shining by calculating the inverse of the measured cloud cover.
The climate risk score is based on a place’s estimated risk of water and heat stress, flooding, hurricanes, typhoons, sea level rise and wildfires at the county level in 2040, according to Moody’s ESG Solutions. The risk levels are relative to global distributions. The scores reflect aggregate risk; a place that is moderately exposed to three or four hazards may have a higher score than a place that is extremely exposed to just one. The analysis does not account for actions that individual counties have taken to prepare for climate hazards.
While some places are at higher risk of physical damage from climate change than others, every place in the United States will be affected in some way, directly or indirectly. There’s no running from climate change.
Politics
The scores for Democrats, Republicans and political diversity are based on research by the political scientists Ryan D. Enos and Jacob Brown. They estimated Americans’ political affiliations by using registered voters’ party identifications, voting history, demographics and precinct-level election results.
To validate their partisanship estimates, Mr. Enos and Mr. Brown asked 10,000 voters which party they belonged to. The estimates matched the voters’ responses 77 percent of the time.
Schools
The school quality score is based on data from Niche. To produce its school grades, Niche uses a range of data, including graduation and absenteeism rates, test scores, money spent per student and survey responses from students and parents.
The score for each place refers to the quality of the public schools that serve it, weighted by those schools’ number of students.
Crime
The crime score is based on relative crime risk, according to Applied Geographic Solutions. The company uses city-, county- and state-level crime reports from the F.B.I., along with socioeconomic variables from the census, to create a model that estimates each place’s risk of crime relative to the national average. No variables related to race, ethnicity, ancestry or language spoken at home were used to create the model.
Violent and nonviolent crime have the same weight in the final score calculation. A place that has the same crime risk as the national average receives a score of 5.
The score provides an estimate of crime, not safety, which is more subjective. Someone who grew up in a town where people don’t lock their doors might think that New York City is much more dangerous than someone who grew up in Brooklyn might.
A useful way to think about the crime score is to search for places that you’ve lived in. Compare their scores to how safe you felt in those places and use those values as benchmarks.
Commute
The commute score is based on the share of workers over the age of 16 whose daily commute was under 30 minutes, according to the census.
Population density
A simple way to calculate population density is to divide each place’s population by its area.
In some places, though, people are unevenly distributed. Anchorage, for example, has an area of more than 1,900 square miles — about the size of Delaware. But most of Anchorage’s land is covered by parkland. If we used a simple density calculation, Alaska’s most populous city would rank among the least densely populated places in the United States.
To sidestep differences in population distribution within places, we calculated the population density of every populated block group (one of the smaller areas that the census defines) in a place. Then we found the population-weighted average density across those block groups. The final density score captures the density observed by the average person in a place.
Space for money
The space for money score is based on the median price per square foot of housing listings posted on Realtor.com over the past two years. We included data only for places with 30 or more listings. The score does not account for the sold price of houses.
Mountains
The mountains score is based on the share of the area within 60 miles of a place’s center that is covered by mountains, with taller mountains counting more. To define what counts as a mountain, we used classifications from “Modeling Global Hammond Landform Regions From 250-m Elevation Data” by Deniz Karagulle, Charlie Frye, Roger Sayre, Sean Breyer, Peter Aniello, Randy Vaughan and Dawn Wright.
Because we’re using a 60-mile buffer for each place’s denominator, places on islands and coastal areas — even ones that are at the base of a mountain, like Honolulu — sometimes receive low scores because they are mostly surrounded by water.
Trees
The tree coverage score is based on the share of the area within a 30-minute drive from a place’s center that is covered by tree canopy, according to EarthDefine’s U.S. Tree Map.
Air quality
The air quality score is based on a place’s median Air Quality Index in 2020, as measured by the Environmental Protection Agency. Air pollutant levels are very low across the United States; most places are safe even for people with respiratory conditions, and therefore have an air quality score of 10 out of 10.
Because we’re using the median Air Quality Index, places that experience occasional wildfires or seasonal smog can still have high air quality scores.
The Air Quality Index is available for a selection of counties and core-based statistical areas, but it is missing for many of the places in our quiz.
Amenities
The scores for restaurants, music venues and gay bars are based on data from Yelp, which provided business counts by place. For each place A, we added the number of businesses in it to the number of businesses in places B, C and D that fell within a 15- or 30-minute drive from A’s center, weighted by the share of B’s, C’s and D’s areas that overlapped with the driving range from A’s center. That way, suburbs received points for amenities in nearby cities.
The restaurant score is based on how many types of restaurants a place has or how many types of restaurants are within a 30-minute drive from a place’s center — using whichever number is larger.
The music venue and gay bar scores measure the number of those businesses a place has, along with the number of those businesses within a 15-minute drive from its center. The gay bar counts also include lesbian bars.
We included the gay bar score to help readers find places with L.G.B.T.Q. people. It’s imperfect; there are places with relatively large L.G.B.T.Q. populations but few or no gay bars. The census has begun to collect better data about same-sex relationships, but those statistics weren’t yet available for most of our places.
Transgender rights
The transgender rights score is based on whether a place’s state has enacted policies that support transgender residents as of July 25, according to the Human Rights Campaign.
If a state excludes transition-related health care coverage from Medicaid or limits transgender students’ participation in athletics, it receives a score of zero. States can receive a score up to 10 based on the degree to which they outlaw discrimination based on gender identity in housing and public accommodations and whether they explicitly include transition-related health care coverage in their Medicaid policies.
Because the score is based on state-level policies, it does not capture differences in attitudes toward transgender people within each state.
Abortion rights
The abortion rights score is based on the abortion-related policies of a place’s state, according to the Guttmacher Institute, which tracks policies such as whether a state excludes abortions from Medicaid coverage, requires parental involvement for abortions or protects abortion rights.
Places can lose points if they are far from an abortion facility or if those facilities require two visits before they will perform an abortion. The distance analysis uses driving times from the population center of each place and a list of facilities that were open as of June 1. The facilities list was provided by Caitlin Myers, an economics professor at Middlebury College.
Jobs
The jobs score is based on a place’s unemployment rates and median household income, according to the census, as well as county- and metro-level job listings in April through August 2021 from Emsi Burning Glass, via Opportunity Insights. The score also accounts for state-level venture capital investments in 2020, provided by the National Venture Capital Association.
Because the jobs score factors in median household income, it is biased toward places where people with high-paying jobs live — not necessarily work. As a result, some suburbs may have higher scores than the cities that they are near.
Income mobility
The income mobility score is based on the mean household income rank of people ages 31 to 37 who grew up in families in the bottom 25 percent of income distribution, as calculated by Opportunity Atlas. Income is measured as mean earnings in 2014 and 2015, using county-level data. Data is mapped to areas where children grew up, regardless of where they lived as adults.
The metric is based on the adult incomes of children born from 1978 to 1983. The researchers at Opportunity Atlas write that on the whole, “historical outcome data prove to be much better predictors of outcomes than more recent data on poverty rates or test scores.” Still, places where the economy boomed and then bottomed out, like some oil and fracking towns, may have surprisingly high scores when compared to those places’ present-day economies.
Health care
The health care score is based on a county-level rank that measures health care system capacity and resources (in terms of work force and infrastructure), health care costs, per capita spending and public funding, according to selected metrics used in the Community Community Vulnerability Index and Covid-19 Vaccine Coverage Index. Data was collected and provided by Surgo Ventures.