COGS 108 - Stepwise Health: Examining the Relationship Between Walkability and Physical Health in Counties and States

Permissions

Place an X in the appropriate bracket below to specify if you would like your group's project to be made available to the public. (Note that student names will be included (but PIDs will be scraped from any groups who include their PIDs).

• [ X ] YES - make available
• [ ] NO - keep private

Names

• Dashan McCain
• Luke Santa-Presca
• Leah Seseri
• Ya-Lun Wu
• Kenny Zhang

Abstract

This study investigates the relationship between walkability and health problems at both county and state levels in the United States. Building upon existing literature that highlights the importance of walkable environments, we explore whether states with higher walkability scores exhibit lower levels of health problems and improved health outcomes.

Drawing from publicly available datasets, including a walkability index across cities and states, physical health data rankings across counties and states, and a Core-Based Statistical Areas (CBSAs) county to city dataset, this study employs a multi-faceted approach to data analysis. By merging and cleaning these datasets, we construct comprehensive frameworks to assess the relationship between walkability and health outcomes across different geographic scales.

Our findings suggest a nuanced relationship between walkability and health indicators. While states with higher walkability scores generally exhibit lower levels of physical inactivity, the correlation is influenced by various socio-economic and environmental factors. Through visualizations and statistical analyses, this study highlights the importance of considering intersectional factors in promoting walkable communities and public health interventions.

Research Question

What is the relationship between walkability and health problems at the county and state level? Specifically, we will examine health problems associated with a lack of physical exercise, particularly those attributable to 'lack of walking'.

Background and Prior Work

For many young individuals, determining their future while navigating the complexities of academic and social life often take precedence over health concerns. As we grapple with adulthood, a recurring issue among college students is the dilemma of transportation. Undergraduates who have cars struggle with the costs that come from parking or gas, while other students are left with the option of public transportation which is often time-consuming. The seemingly simple solution to this is to walk, however, this is hindered due to the unwalkable nature of certain places. Thus, we began to ponder the varying degrees of walkability in cities we have visited or lived in, and the subjective criteria we use to define a location as walkable. Recognizing walking as a form of physical activity, we wanted to investigate whether states we deemed more walkable were associated with a healthier population.

A reference¹ we found online is from the Boston University School of Public Health. They searched for links between city walkability and BMI levels, which is very similar to our experiment. Their findings supported the conclusion that there is a correlation between the walkability of state and BMI levels. But they also found there is a distinction between white and nonwhite people. This disparity has been "borne from systemic racism and policies that have created barriers for many communities of color to embrace health-protective behaviors." When we drafted the project, we did not have this in mind, but we will investigate this to see if we also reached the same conclusions.

Another reference² we found is from the National Library of Medicine. In addition, they investigated whether there was a link between walkable communities and physical and social health in their Texas town. They discovered a positive correlation between the two, as well as statistics indicating that inhabitants are driving less. However, they admit that their conclusions were based on survey data, which limited the number of people who responded to the survey depending on characteristics such as computer and internet availability, as well as overall interest in the study.

We intend to carry out a thorough analysis using publicly available census data from several locations in order to set our approach apart from previous research. Although walkability, health outcomes, and demographic characteristics have all been studied before, our focus on analyzing a wide variety of urban environments will offer a more thorough and nuanced understanding of these correlations. We hope to capture a wider range of demographics by using census data, taking into account elements like population density, racial variety, and socioeconomic position.

1. ^ Morris, Vivien. “US Neighborhood Walkability Influences Physical Activity, BMI Levels.” SPH US Neighborhood Walkability Influences Physical Activity BMI Levels Comments, 2 Feb. 2023, https://www.bu.edu/sph/news/articles/2023/us-neighborhood-walkability-influences-physical-activity-bmi-levels/.
2. ^ Zhu, Xuemei, et al. “Walkable Communities: Impacts on Residents’ Physical and Social Health: Researchers from Texas A&M University Studied Residents in a Newly Developed ‘Walkable Community’ in Austin, Texas to See How It Changed Their Habits for Physical Activity and Whether It Increased Social Interaction and Cohesion in the Community.” World Health Design, U.S. National Library of Medicine, July 2013, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8776244/.

Hypothesis

Our main hypothesis is that states that have a higher walkability score will have a lower level of health problems since the increase in walkability may urge residents of the state to walk more throughout their day. Not only will the availability of everyday stores be closer by walking, the residents may be more inclined to walk/bike to work instead of driving.

Data

Data overview

• Dataset #1
• Dataset Name: Walkability Index
• Link to the dataset: https://catalog.data.gov/dataset/walkability-index1
• Number of observations: 220740
• Number of variables: 117

The walkability dataset contains data that ranks block groups according to their relative walkability and assigns each block group a walkability index score using the formula (w/3) + (x/3) + (y/6) + (z/6) where w = block group's ranked score for intersection density, x = block group's ranked score for proximity to transit stops, y = block group's ranked score for employment mix, and z = block group's ranked score for employment and household mix.

The dataset's important variables include 'STATEFP', 'CBSA_Name', and 'NatWalkInd'. 'STATEFP' contains integers that point to a state, 'CBSA_Name' contains strings of various cities in the United States, and 'NatWalkInd' contains an integer up to 20, the maximum possible score for a city's walkability.

In order to clean the walkability dataset, we needed to remove all columns that were not relevant to this project, remove entries from the District of Columbia and Puerto Rico because they are not states, calculate the mean of all NatWalkInd because there are multiple entires for cities, and create a new dataframe with all of the changes.

• Dataset #2
• Dataset Name: County Health Rankings
• Link to the dataset: https://www.countyhealthrankings.org/sites/default/files/media/document/analytic_data2021.csv
• Number of observations: 3195
• Number of variables: 690

The health dataset contains health information about states in the US broken down by county. The dataset contains a large amount of measures so we must focus on those that directly affect a person’s physical health.

The important variables in this dataset are ‘State Abbreviation’, ‘Name’ which contains strings representing counties for each state, ‘Poor or fair health raw value’ which contains a float representing the percentage of adults who record poor/fair health, ‘Poor physical health days raw value’ which contains a float that represents the number of poor physical health days, ‘Physical inactivity raw value’ which contains a float representing the percentage of adults who report no leisure-time physical activity, ‘Frequent physical distress raw value’ which contains a float representing the proportion of adults who have frequent physical distress, ‘Median household income raw value’ contains an integer representing the median income in a household and is broken down by ethnicity, ‘Percentage of households with high housing costs’ represents a float showing the percent of households that pay a large amount for housing, percent of each ethnicity as a float, and ‘Population raw value’ showing the amount of people that live in each state.

In order to clean the health dataset, we only included the columns that measured physical health, income, and ethnicity. We got the top two most walkable states and bottom two least walkable states along with California from the walkability dataset and used the ‘Name’ variable to get only these five states and remove the rows of counties. Then, we created a new dataframe with the rows in order of most walkable to least walkable.

• Dataset #3
• Dataset Name: Census Core-Based Statistical Area (CBSA) to Federal Information Processing Series (FIPS) County Crosswalk
• Link to the dataset: https://www.nber.org/research/data/census-core-based-statistical-area-cbsa-federal-information-processing-series-fips-county-crosswalk
• Number of observations: 1915
• Number of variables: 12

This dataset has many additional features, however we will only use the columns Core-Based Statistical Area (CBSA) and county equivalent. This will essentially assist us in matching city names to their corresponding counties, allowing us to study the national walking index in relation to its respective county, resulting in a more thorough analysis. The remaining cities that the dataset was unable to match to a county will be matched manually.

To clean the county dataset, we needed to eliminate any columns that were irrelevant to our project and generate a new DataFrame with all of the adjustments.

We aim to use these three datasets, thus the merging technique will be to assign the NatWalkInd data to the appropriate county or state and construct two separate DataFrames.

Walkability Dataset

import pandas as pd

walkability_df = pd.read_csv('walkability.csv')

walkability_df = walkability_df[walkability_df['STATEFP'] != 11]
walkability_df = walkability_df[walkability_df['STATEFP'] != 72] 

selected_columns = ['STATEFP', 'CBSA_Name', 'NatWalkInd']
walkability_df = walkability_df[selected_columns]

walkability_df

state_mapping = {
    '1': 'ALABAMA',
    '2': 'ALASKA',
    '4': 'ARIZONA',
    '5': 'ARKANSAS',
    '6': 'CALIFORNIA',
    '8': 'COLORADO',
    '9': 'CONNECTICUT',
    '10': 'DELAWARE',
    '12': 'FLORIDA',
    '13': 'GEORGIA',
    '15': 'HAWAII',
    '16': 'IDAHO',
    '17': 'ILLINOIS',
    '18': 'INDIANA',
    '19': 'IOWA',
    '20': 'KANSAS',
    '21': 'KENTUCKY',
    '22': 'LOUISIANA',
    '23': 'MAINE',
    '24': 'MARYLAND',
    '25': 'MASSACHUSETTS',
    '26': 'MICHIGAN',
    '27': 'MINNESOTA',
    '28': 'MISSISSIPPI',
    '29': 'MISSOURI',
    '30': 'MONTANA',
    '31': 'NEBRASKA',
    '32': 'NEVADA',
    '33': 'NEW HAMPSHIRE',
    '34': 'NEW JERSEY',
    '35': 'NEW MEXICO',
    '36': 'NEW YORK',
    '37': 'NORTH CAROLINA',
    '38': 'NORTH DAKOTA',
    '39': 'OHIO',
    '40': 'OKLAHOMA',
    '41': 'OREGON',
    '42': 'PENNSYLVANIA',
    '44': 'RHODE ISLAND',
    '45': 'SOUTH CAROLINA',
    '46': 'SOUTH DAKOTA',
    '47': 'TENNESSEE',
    '48': 'TEXAS',
    '49': 'UTAH',
    '50': 'VERMONT',
    '51': 'VIRGINIA',
    '53': 'WASHINGTON',
    '54': 'WEST VIRGINIA',
    '55': 'WISCONSIN',
    '56': 'WYOMING'
}

for state_code, state_name in state_mapping.items():
    walkability_df.loc[walkability_df['STATEFP'] == int(state_code), 'CBSA_Name'] = walkability_df['CBSA_Name'].fillna(state_name)

walkability_df = walkability_df.dropna()
walkability_df

state = (
    walkability_df.groupby('STATEFP')
    .agg({'NatWalkInd': 'mean'})
    .reset_index()
    .sort_values(by='NatWalkInd', ascending=False)
)

state['STATEFP'] = state['STATEFP'].astype(str)
state['STATE'] = state['STATEFP'].map(state_mapping)
state.drop(columns=['STATEFP'], inplace=True)
state.rename(columns={'STATE': 'State'}, inplace=True)
state = state[['State', 'NatWalkInd']]

state.reset_index(drop=True, inplace=True) 
state

County Dataset

county_df = pd.read_csv('county.csv')

county = (
    walkability_df.groupby(['STATEFP', 'CBSA_Name'])
    .agg({'NatWalkInd': 'mean'})
    .reset_index()
    .sort_values(by='NatWalkInd', ascending=False)
)

city_to_county_mapping = {
    'Albany-Lebanon, OR': 'Linn County',
    'Atlanta-Sandy Springs-Alpharetta, GA': ['Barrow County', 'Butts County', 'Carroll County', 'Clayton County', 
                                             'Coweta County', 'Dawson County', 'DeKalb County', 'Douglas County', 
                                             'Fayette County', 'Forsyth County', 'Fulton County', 'Gwinnett County', 
                                             'Heard County', 'Henry County', 'Jasper County', 'Lumpkin County', 
                                             'Meriwether County', 'Morgan County', 'Newton County', 'Pickens County', 
                                             'Pike County','Rockdale County', 'Spalding County', 'Walton County'],
    'Ashtabula, OH': 'Ashtabula County',
    'Atmore, AL': 'Escambia County',
    'Austin-Round Rock-Georgetown, TX': ['Bastrop County', 'Caldwell County', 'Hays County', 'Travis County', 
                                         'Williamson County'],
    'Bakersfield, CA': 'Kern County',
    'Bardstown, KY': 'Nelson County',
    'Bennettsville, SC': 'Marlboro County',
    'Berlin, NH': 'Coos County',
    'Big Stone Gap, VA': 'Wise County',
    'Birmingham-Hoover, AL': ['Bibb County', 'Blount County', 'Chilton County', 'Coosa County', 'Cullman County', 
                              'Jefferson County', 'St. Clair County', 
                              'Shelby County', 'Talladega County', 'Walker County'],
    'Blacksburg-Christiansburg, VA': ['Floyd County', 'Giles County', 'Montgomery County', 'Pulaski County', 
                                      'Radford city'],
    'Bridgeport-Stamford-Norwalk, CT': 'Fairfield County',
    'Brownsville, TN': 'Haywood County',
    'Brunswick, GA': ['Brantley County', 'Glynn County', 'McIntosh County'],
    'Bucyrus-Galion, OH': 'Crawford County',
    'California-Lexington Park, MD': ['Calvert County', "St. Mary's County"],
    'Carbondale-Marion, IL': ['Jackson County', 'Williamson County'],
    'Chambersburg-Waynesboro, PA': 'Franklin County',
    'Central City, KY': 'Muhlenberg County',
    'Chicago-Naperville-Elgin, IL-IN-WI': ['Cook County', 'DeKalb County', 'DuPage County', 'Grundy County', 
                                           'Kane County', 'Kendall County', 'Lake County', 'McHenry County',
                                           'Will County', 'Jasper County', 'Lake County', 'Newton County', 
                                           'Porter County'],
    'Cleveland-Elyria, OH': ['Ashtabula County', 'Cuyahoga County', 'Geauga County', 'Lake County', 'Lorain County', 
                             'Medina County'],
    'Coffeyville, KS': 'Montgomery County',
    'Coos Bay, OR': 'Coos County',
    'Craig, CO': 'Moffat County',
    'Cullowhee, NC': 'Jackson County',
    'Dayton, TN': 'Rhea County',
    'Dayton-Kettering, OH': ['Greene County', 'Miami County', 'Montgomery County'],
    'Denver-Aurora-Lakewood, CO': ['Adams County', 'Arapahoe County', 'Broomfield County', 'Clear Creek County', 
                                   'Denver County', 'Douglas County', 'Elbert County',
                                   'Gilpin County', 'Jefferson County', 'Park County'],
    'Elizabethtown-Fort Knox, KY': ['Hardin County', 'Larue County'],
    'Evanston, WY': ['Rich County', 'Uinta County'],
    'Evansville, IN-KY': ['Posey County', 'Vanderburgh County', 'Warrick County'],
    'Fairbanks, AK': 'Fairbanks North Star Borough',
    'Fairfield, IA': 'Jefferson County',
    'Fernley, NV': 'Lyon County',
    'Fort Collins, CO': 'Larimer County',
    'Fort Madison-Keokuk, IA-IL-MO': 'Lee County',
    'Fort Polk South, LA': 'Vernon Parish County',
    'Gardnerville Ranchos, NV': ['Alpine County', 'Douglas County'],
    'Georgetown, SC': 'Georgetown County',
    'Glenwood Springs, CO': 'Garfield County',
    'Grand Rapids-Kentwood, MI': ['Barry County', 'Ionia County', 'Kent County', 'Montcalm County', 'Muskegon County', 
                                  'Ottawa County', 'Itasca County'],
    'Grants, NM': 'Cibola County',
    'Greenville-Anderson, SC': ['Greenville County', 'Anderson County', 'Laurens County', 'Pickens County'],
    'Hartford-East Hartford-Middletown, CT': ['Hartford County', 'Litchfield County', 'Middlesex County', 
                                              'New London County', 'Tolland County', 'Windham County'],
    'Helena-West Helena, AR': 'Phillips County',
    'Hilo, HI': 'Hawaii County',
    'Hilton Head Island-Bluffton, SC': ['Beaufort County', 'Jasper County'],
    'Hope, AR': 'Hempstead County',
    'Houma-Thibodaux, LA': ['Lafourche Parish', 'Terrebonne Parish'],
    'Houston-The Woodlands-Sugar Land, TX': ['Austin County', 'Brazoria County', 'Fort Bend County', 'Galveston County', 
                                             'Harris County', 'Liberty County', 'Montgomery County', 
                                             'San Jacinto County', 'Waller County'],
    'Indianapolis-Carmel-Anderson, IN': ['Boone County', 'Brown County', 'Hamilton County', 'Hancock County', 
                                         'Hendricks County', 'Johnson County', 'Madison County', 'Marion County', 
                                         'Morgan County', 'Shelby County', 'Tipton County'],
    'Indianola, MS': 'Sunflower County',
    'Jackson, OH': 'Jackson County',
    'Jamestown-Dunkirk-Fredonia, NY': 'Chautauqua County',
    'Jasper, AL': 'Walker County',
    'Jennings, LA': 'Jefferson Davis Parish',
    'Joplin, MO': ['Cherokee County', 'Jasper County', 'Newton County'],
    'Kahului-Wailuku-Lahaina, HI': ['Kalawao County', 'Maui County'],
    'Key West, FL': 'Monroe County',
    'Lamesa, TX': 'Dawson County',
    'Las Vegas-Henderson-Paradise, NV': 'Clark County',
    'Lebanon, NH-VT': ['Grafton County', 'Sullivan County'],
    'Levelland, TX': 'Hockley County',
    'London, KY': 'Laurel County',
    'Longview, WA': 'Cowlitz County',
    'Madera, CA': 'Madera County',
    'Malone, NY': 'Franklin County',
    'Maysville, KY': 'Mason County',
    'Miami-Fort Lauderdale-Pompano Beach, FL': ['Broward County', 'Miami-Dade County', 'Palm Beach County'],
    'Mount Gay-Shamrock, WV': 'Logan County',
    'Muskegon, MI': 'Muskegon County',
    'Myrtle Beach-Conway-North Myrtle Beach, SC-NC': 'Horry County',
    'New Castle, PA': 'Lawrence County',
    'New Haven-Milford, CT': 'New Haven County',
    'New York-Newark-Jersey City, NY-NJ-PA': ['Bergen County', 'Essex County', 'Hudson County', 'Hunterdon County', 
                                              'Middlesex County', 'Monmouth County', 'Morris County','Ocean County', 
                                              'Passaic County', 'Somerset County', 'Sussex County', 'Union County', 
                                              'Bronx County', 'Kings County','Nassau County', 'New York County', 
                                              'Putnam County', 'Queens County', 'Richmond County', 'Rockland County', 
                                              'Suffolk County','Westchester County'],
    'North Port-Sarasota-Bradenton, FL': ['Manatee County', 'Sarasota County'],
    'North Vernon, IN': 'Jennings County',
    'Norwich-New London, CT': 'New London County',
    'Ocean City, NJ': 'Cape May County',
    'Ogden-Clearfield, UT': ['Davis County', 'Morgan County', 'Weber County'],
    'Ogdensburg-Massena, NY': 'St. Lawrence County',
    'Omaha-Council Bluffs, NE-IA': ['Harrison County', 'Mills County', 'Pottawattamie County', 'Cass County', 
                                    'Douglas County', 'Sarpy County', 'Saunders County', 'Washington County'],
    'Panama City, FL': ['Bay County', 'Washington County'],
    'Pearsall, TX': 'Frio County',
    'Pecos, TX': 'Reeves County',
    'Point Pleasant, WV-OH': 'Mason County',
    'Portales, NM': 'Roosevelt County',
    'Poughkeepsie-Newburgh-Middletown, NY': ['Dutchess County', 'Orange County'],
    'Prineville, OR': 'Crook County',
    'Provo-Orem, UT': ['Juab County', 'Utah County'],
    'Racine, WI': 'Racine County',
    'Parsons, KS': 'Labette County',
    'Rockport, TX': 'Aransas County',
    'Salisbury, MD-DE': ['Somerset County', 'Wicomico County'],
    'Salt Lake City, UT': ['Salt Lake County', 'Tooele County'],
    'San Francisco-Oakland-Berkeley, CA': ['Alameda County', 'Contra Costa County', 'Marin County', 'San Francisco County', 
                                           'San Mateo County'],
    'Scottsburg, IN': 'Scott County',
    'Sebastian-Vero Beach, FL': 'Indian River County',
    'Sebring-Avon Park, FL': 'Highlands County',
    'Shelby, NC': 'Cleveland County',
    'Sioux Falls, SD': ['Rock County', 'Lincoln County', 'McCook County', 'Minnehaha County', 'Turner County'],
    'St. Marys, GA': 'Camden County',
    'Staunton, VA': ['Augusta County', 'Staunton city', 'Waynesboro city'],
    'Stevens Point, WI': 'Portage County',
    'Stockton, CA': 'San Joaquin County',
    'The Villages, FL': 'Sumter County',
    'Union, SC': 'Union County',
    'Vineland-Bridgeton, NJ': 'Cumberland County',
    'Virginia Beach-Norfolk-Newport News, VA-NC': ['Camden County', 'Currituck County', 'Gates County', 
                                                   'Gloucester County','Isle of Wight County', 'James City County', 
                                                   'Matthews County','Surry County', 'York County', 
                                                   'Chesapeake city', 'Hampton city', 'Newport News city',
                                                   'Norfolk city', 'Poquoson city', 'Portsmouth city','Suffolk city', 
                                                   'Virginia Beach city', 'Williamsburg city'],
    'Wauchula, FL': 'Hardee County',
    'Wausau-Weston, WI': 'Marathon County',
    'Wenatchee, WA': ['Chelan County', 'Douglas County'],
    'West Point, MS': 'Clay County',
    'Whitewater, WI': 'Walworth County',
    'Winfield, KS': 'Cowley County',
    'Worcester, MA-CT': 'Worcester County',
    'Youngstown-Warren-Boardman, OH-PA': ['Mahoning County', 'Trumbull County'],
}

selected_columns = ['cbsatitle', 'countycountyequivalent']
county_df = county_df[selected_columns]

county = pd.merge(county, county_df, left_on='CBSA_Name', right_on='cbsatitle', how='left')
county.drop(columns=['cbsatitle'], inplace=True)

county['countycountyequivalent'] = county['countycountyequivalent'].fillna(county['CBSA_Name'].map(city_to_county_mapping))
county = county.explode('countycountyequivalent')
county.dropna(inplace=True) 
county.drop('CBSA_Name', axis=1, inplace=True)

county['STATEFP'] = county['STATEFP'].astype(str)
county['STATE'] = county['STATEFP'].map(state_mapping)

desired_order = ['STATE', 'countycountyequivalent', 'NatWalkInd']
county = county[desired_order]
county.rename(columns={'countycountyequivalent': 'County'}, inplace=True)

county.reset_index(drop=True, inplace=True)
county

Health Dataset

health_df = pd.read_csv('health.csv', low_memory=False)

selected_columns = ['State Abbreviation', 'Name', 'Poor or fair health raw value', 
                     'Poor physical health days raw value', 'Physical inactivity raw value', 
                     'Frequent physical distress raw value', 'Median household income raw value', 
                     'Median household income (Asian)', 'Median household income (Black)',
                     'Median household income (Hispanic)', 'Median household income (White)', 
                     'Percentage of households with high housing costs', 
                     '% Non-Hispanic Black raw value', '% American Indian & Alaska Native raw value', 
                     '% Asian raw value', '% Native Hawaiian/Other Pacific Islander raw value', 
                     '% Hispanic raw value', '% Non-Hispanic White raw value','Population raw value']

health_df = health_df[selected_columns]

health_df

Merging Datasets by County and State

exclude_keywords = 'county|borough|area|municipality|city|parish|District of Columbia'
health_by_state = health_df[~health_df['Name'].str.contains(exclude_keywords, case=False)].copy()
health_by_state['Name'] = health_by_state['Name'].str.upper()
health_by_state.rename(columns={'Name': 'State'}, inplace=True)
health_by_state.reset_index(drop=True, inplace=True)
health_by_state

health_by_state = pd.merge(state, health_by_state, on='State')
health_by_state.drop('State Abbreviation', axis=1, inplace=True)
health_by_state

health_by_county = health_df[health_df['Name'].str.contains('county|borough|area|municipality|city|parish', case=False)].copy()
health_by_county.drop(0, inplace=True)
health_by_county.rename(columns={'Name': 'County'}, inplace=True)
health_by_county.reset_index(drop=True, inplace=True)
health_by_county

state_abbreviations_to_names = {
    'AL': 'ALABAMA',
    'AK': 'ALASKA',
    'AZ': 'ARIZONA',
    'AR': 'ARKANSAS',
    'CA': 'CALIFORNIA',
    'CO': 'COLORADO',
    'CT': 'CONNECTICUT',
    'DE': 'DELAWARE',
    'FL': 'FLORIDA',
    'GA': 'GEORGIA',
    'HI': 'HAWAII',
    'ID': 'IDAHO',
    'IL': 'ILLINOIS',
    'IN': 'INDIANA',
    'IA': 'IOWA',
    'KS': 'KANSAS',
    'KY': 'KENTUCKY',
    'LA': 'LOUISIANA',
    'ME': 'MAINE',
    'MD': 'MARYLAND',
    'MA': 'MASSACHUSETTS',
    'MI': 'MICHIGAN',
    'MN': 'MINNESOTA',
    'MS': 'MISSISSIPPI',
    'MO': 'MISSOURI',
    'MT': 'MONTANA',
    'NE': 'NEBRASKA',
    'NV': 'NEVADA',
    'NH': 'NEW HAMPSHIRE',
    'NJ': 'NEW JERSEY',
    'NM': 'NEW MEXICO',
    'NY': 'NEW YORK',
    'NC': 'NORTH CAROLINA',
    'ND': 'NORTH DAKOTA',
    'OH': 'OHIO',
    'OK': 'OKLAHOMA',
    'OR': 'OREGON',
    'PA': 'PENNSYLVANIA',
    'RI': 'RHODE ISLAND',
    'SC': 'SOUTH CAROLINA',
    'SD': 'SOUTH DAKOTA',
    'TN': 'TENNESSEE',
    'TX': 'TEXAS',
    'UT': 'UTAH',
    'VT': 'VERMONT',
    'VA': 'VIRGINIA',
    'WA': 'WASHINGTON',
    'WV': 'WEST VIRGINIA',
    'WI': 'WISCONSIN',
    'WY': 'WYOMING'
}

health_by_county['STATE'] = health_by_county['State Abbreviation'].map(state_abbreviations_to_names)
health_by_county.drop('State Abbreviation', axis=1, inplace=True)

state_column = health_by_county['STATE']
health_by_county.drop('STATE', axis=1, inplace=True)
health_by_county.insert(0, 'STATE', state_column)

health_by_county

health_by_county = pd.merge(county, health_by_county, on=['STATE', 'County'], how='outer')
health_by_county.dropna(subset=['NatWalkInd'], inplace=True)

health_by_county

health_by_ca = health_by_county[health_by_county['STATE'] == 'CALIFORNIA'].copy()
health_by_ca.sort_values(by='NatWalkInd', ascending=False, inplace=True)
health_by_ca

Results

Exploratory Data Analysis

The National Walking Index (NatWalkInd) is employed as a statistic in this study to investigate the association between walkability and health at the county and state level. The first bar graphs, which display NatWalkInd scores for each state in the United States, provide the foundation for further investigation. Scatter plots and joint plots reveal a potential negative relationship between walkability and physical inactivity, with a modest trend indicating that better walkability is related with reduced physical inactivity. The report also examines the counties in each state that have the greatest and poorest health scores. It demonstrates a strong relationship between walkability and overall health, lending support to the notion that a more walkable environment may lead to better health outcomes.

Simple Graph Visualizations

import matplotlib.pyplot as plt
import seaborn as sns

plt.figure(figsize=(15, 8))

sns.barplot(data=health_by_state, x='State', y='NatWalkInd', color='skyblue', dodge=True)

plt.title('Bar Graph of National Walking Index by State')
plt.xticks(rotation=90)
plt.show()

These are the fifty US states, graded from best to worst in terms of walkability. The index score is calculated using this formula: ((ranked score for intersection density) / 3) + ((ranked score for closeness to transit stations) / 3) + ((ranked score for employment mix) / 6) + ((ranked score for employment and household mix) / 6). The division shows that certain elements are more important than others in deciding the final score.

plt.figure(figsize=(15, 8))

sns.barplot(data=health_by_ca, x='County', y='NatWalkInd', color='skyblue', dodge=True)

plt.title('Bar Graph of National Walking Index by CA County')
plt.xticks(rotation=90)
plt.show()

This is the same graph, but for the state of California. Glad to see San Diego County have a high walkable score.

Exploring the Link: Walkability and its Impact on Health

fig, axes = plt.subplots(1, 2, figsize=(12, 6))

health_by_ca_sorted = health_by_ca.sort_values(by='Physical inactivity raw value', ascending=True)
health_by_ca_sorted['Physical inactivity raw value'] = pd.to_numeric(
    health_by_ca_sorted['Physical inactivity raw value'], errors='coerce')
sns.scatterplot(x='NatWalkInd', y='Physical inactivity raw value', data=health_by_ca_sorted, ax=axes[0])
sns.regplot(x='NatWalkInd', y='Physical inactivity raw value', data=health_by_ca_sorted, scatter=False, ax=axes[0])
axes[0].set_xlabel('NatWalkInd')
axes[0].set_ylabel('Physical inactivity raw value')
axes[0].set_title('Scatter plot of NatWalkInd vs Physical inactivity raw value')

health_by_state_sorted = health_by_state.sort_values(by='Physical inactivity raw value', ascending=True)
health_by_state_sorted['Physical inactivity raw value'] = pd.to_numeric(
    health_by_state_sorted['Physical inactivity raw value'], errors='coerce')
sns.scatterplot(x='NatWalkInd', y='Physical inactivity raw value', data=health_by_state_sorted, ax=axes[1])
sns.regplot(x='NatWalkInd', y='Physical inactivity raw value', data=health_by_state_sorted, scatter=False, ax=axes[1])
axes[1].set_xlabel('NatWalkInd')
axes[1].set_ylabel('Physical inactivity raw value')
axes[1].set_title('Scatter plot of NatWalkInd vs Physical inactivity raw value')

plt.tight_layout()
plt.show()

The left graph compares California counties' walkability and physical inactivity. The right uses the same comparisons, but for all fifty states. As you can see, they are pretty comparable, however for California, we believe the trend line is misleading in pointing downward because the data is highly spread out. For all fifty states, we believe the trend line is correct and shows a declining tendency. A decreasing trend indicates that the more walkable a state is, the more people are physically active.

fig, axes = plt.subplots(1, 2, figsize=(12, 6))

health_by_ca_sorted = health_by_ca.sort_values(by='Poor or fair health raw value', ascending=True)
health_by_ca_sorted['Poor or fair health raw value'] = pd.to_numeric(
    health_by_ca_sorted['Poor or fair health raw value'], errors='coerce')
sns.scatterplot(x='NatWalkInd', y='Poor or fair health raw value', data=health_by_ca_sorted, ax=axes[0])
sns.regplot(x='NatWalkInd', y='Poor or fair health raw value', data=health_by_ca_sorted, scatter=False, ax=axes[0])
axes[0].set_xlabel('NatWalkInd')
axes[0].set_ylabel('Poor or fair health raw value')
axes[0].set_title('Scatter plot of NatWalkInd vs Poor or fair health raw value')

health_by_state_sorted = health_by_state.sort_values(by='Poor or fair health raw value', ascending=True)
health_by_state_sorted['Poor or fair health raw value'] = pd.to_numeric(
    health_by_state_sorted['Poor or fair health raw value'], errors='coerce')
sns.scatterplot(x='NatWalkInd', y='Poor or fair health raw value', data=health_by_state_sorted, ax=axes[1])
sns.regplot(x='NatWalkInd', y='Poor or fair health raw value', data=health_by_state_sorted, scatter=False, ax=axes[1])
axes[1].set_xlabel('NatWalkInd')
axes[1].set_ylabel('Poor or fair health raw value')
axes[1].set_title('Scatter plot of NatWalkInd vs Poor or fair health raw value')

plt.tight_layout()
plt.show()

The left graph compares California counties' walkability and poor or fair health. The right uses the same comparisons, but for all fifty states. As you can see, they are pretty comparable, but we do not feel either of these graphs shows a falling tendency. The information is overly spread. We believe that the data varies depending on a number of factors, including access to healthcare, socioeconomic status, education level, environmental impact, public health policies, physical activity rates, smoking and tobacco use, diet and nutrition, chronic disease prevalence, and water and air quality.

health_by_county_sorted = health_by_county.sort_values(by='Physical inactivity raw value', ascending=True)
health_by_county_sorted['Physical inactivity raw value'] = pd.to_numeric(health_by_county_sorted['Physical inactivity raw value'], errors='coerce')
sns.jointplot(x='NatWalkInd', y='Physical inactivity raw value', data=health_by_county_sorted, kind = "reg", dropna = True)
plt.xlabel('NatWalkInd')
plt.ylabel('Physical inactivity raw value')
plt.show()

Since we determined that poor or fair health is not a trustworthy factor to include, the graph compares walkability to physical inactivity in counties across all 50 states. We see a slight downward trend, indicating that the more walkable the area, the less likely physical inactivity is. We say slight because there are more factors contributing to physical inactivity than just how walkable a county or state is.

health_by_county_sorted = health_by_county.sort_values(by='Physical inactivity raw value', ascending=True)
health_by_county_sorted['Physical inactivity raw value'] = pd.to_numeric(
    health_by_county_sorted['Physical inactivity raw value'], errors='coerce')
sns.jointplot(x='NatWalkInd', y='Physical inactivity raw value', data=health_by_county_sorted, kind='hex', gridsize=20, marginal_kws=dict(bins=30))
plt.xlabel('NatWalkInd')
plt.ylabel('Physical inactivity raw value')
plt.show()

This is the same graph, but with a joint plot, which is ideal for visualizing where the data is most significant. Aside from the fact that this is an intriguing graph, we can see that the data is most dense around walkability index scores 6-8, with physical inactivity raw values ranging from 0.23 to 0.32.

Breaking graphs further down

top_health_by_county = health_by_county[health_by_county['STATE'].isin(['RHODE ISLAND', 'CALIFORNIA', 'NEW JERSEY'])]
bot_health_by_county = health_by_county[health_by_county['STATE'].isin(['MISSISSIPPI', 'WEST VIRGINIA', 'ARKANSAS   '])]

plt.figure(figsize=(12, 6))
plt.subplot(1, 2, 1)

top_health_by_county_sorted = top_health_by_county.sort_values(by='Physical inactivity raw value', ascending=True)
top_health_by_county_sorted['Physical inactivity raw value'] = pd.to_numeric(
    top_health_by_county_sorted['Physical inactivity raw value'], errors='coerce')
sns.scatterplot(x='NatWalkInd', y='Physical inactivity raw value', data=top_health_by_county_sorted)
sns.regplot(x='NatWalkInd', y='Physical inactivity raw value', data=top_health_by_county_sorted, scatter=False)
plt.xlabel('NatWalkInd')
plt.ylabel('Physical inactivity raw value')
plt.title('Top Counties: NatWalkInd vs Physical Inactivity')


plt.subplot(1, 2, 2)
bot_health_by_county_sorted = bot_health_by_county.sort_values(by='Physical inactivity raw value', ascending=True)
bot_health_by_county_sorted['Physical inactivity raw value'] = pd.to_numeric(
    bot_health_by_county_sorted['Physical inactivity raw value'], errors='coerce')
sns.scatterplot(x='NatWalkInd', y='Physical inactivity raw value', data=bot_health_by_county_sorted)
sns.regplot(x='NatWalkInd', y='Physical inactivity raw value', data=bot_health_by_county_sorted, scatter=False)
plt.xlabel('NatWalkInd')
plt.ylabel('Physical inactivity raw value')
plt.title('Bottom Counties: NatWalkInd vs Physical Inactivity')
plt.tight_layout()

plt.show()

Because it was difficult to discern whether a high walkability score indicated improved physical health, we compared the top three walkable states to the worst three walkable states. The left graph compares the top counties' walkability and physical inactivity. The comparisons on the right are the same, but for the bottom counties' walkability and physical inactivity. As you can see, they are quite comparable, however we do not believe either of these graphs indicates a downward trend. The information is widely disseminated. Because of the variety of data points, it is possible to argue whether or not a county's walkability reduces physical inactivity.

plt.figure(figsize=(12, 6))
plt.subplot(1, 2, 1)

top_health_by_county_sorted = top_health_by_county.sort_values(by='Poor or fair health raw value', ascending=True)
top_health_by_county_sorted['Poor or fair health raw value'] = pd.to_numeric(
    top_health_by_county_sorted['Poor or fair health raw value'], errors='coerce')
sns.scatterplot(x='NatWalkInd', y='Poor or fair health raw value', data=top_health_by_county_sorted)
sns.regplot(x='NatWalkInd', y='Poor or fair health raw value', data=top_health_by_county_sorted, scatter=False)
plt.xlabel('NatWalkInd')
plt.ylabel('Poor or fair health raw value')
plt.title('Top Counties: NatWalkInd vs Poor or Fair Health')


plt.subplot(1, 2, 2)

bot_health_by_county_sorted = bot_health_by_county.sort_values(by='Poor or fair health raw value', ascending=True)
bot_health_by_county_sorted['Poor or fair health raw value'] = pd.to_numeric(
    bot_health_by_county_sorted['Poor or fair health raw value'], errors='coerce')
sns.scatterplot(x='NatWalkInd', y='Poor or fair health raw value', data=bot_health_by_county_sorted)
sns.regplot(x='NatWalkInd', y='Poor or fair health raw value', data=bot_health_by_county_sorted, scatter=False)
plt.xlabel('NatWalkInd')
plt.ylabel('Poor or fair health raw value')
plt.title('Bottom Counties: NatWalkInd vs Poor or Fair Health')
plt.tight_layout()

plt.show()

Even though we claimed that poor or good health should not be used to compare a county's walkability, the graphs we created were pretty shocking. The left graph compares the top counties' walkability to their poor or fair health. The comparisons on the right are identical, but for the lowest counties' walkability and poor or good health. In terms of health, the top three walkable states have a huge downward trend over the bottom three. The data for the bottom three walkable states is clumped together in the center. Of course, there are several elements that influence someone's health, but it is impossible to dispute that a state's walkability has an impact on health.

Ethics & Privacy

• Are there any biases/privacy/terms of use issues with the data you proposed?
• Biases may include data from different socioeconomic statuses, while terms of use restrict how the data can be used and shared.
• Are there potential biases in your dataset(s), in terms of who it composes, and how it was collected, that may be problematic in terms of it allowing for equitable analysis? (For example, does your data exclude particular populations, or is it likely to reflect particular human biases in a way that could be a problem?)
• Potential biases may include factors such as access to healthcare, socioeconomic status, and cultural differences in reporting health conditions, which could influence health data. Similarly, walkability indices may rely on criteria that do not fully capture the lived experiences of all communities.
• How will you set out to detect these specific biases before, during, and after/when communicating your analysis?
• We conduct literature review and validate the data before analysis. During analysis, we remain sensitive to potential biases and employ exploratory data analysis (EDA) techniques to mitigate their impact. After analysis, we ensure transparency by openly communicating all aspects of the data and our findings.
• Are there any other issues related to your topic area, data, and/or analyses that are potentially problematic in terms of data privacy and equitable impact?
• We need to ensure that the data analyzed for research purposes are anonymized and do not contain personally identifiable information.
• How will you handle issues you identified? (1-2 paragraphs)
• To address the personal privacy concerns identified, we will implement several measures. Firstly, we will remove direct identifiers such as names, addresses, and contact information from the dataset, replacing them with unique identifiers to protect individuals' identities. Additionally, we will employ robust data security measures to safeguard against unauthorized access to sensitive health information. Finally, at the conclusion of the research, we will provide clear and transparent disclosure regarding how participants' data was used and protected throughout the study.

Discusison and Conclusion

In this project, our aim was to investigate the impact of a state or county’s walkability on an individual’s health. Specifically, we wanted to answer the question of whether increased walkability correlates to improved health, possible due to higher levels of physical activity. To address this inquiry, we analyzed three distinct datasets consisting of information on states walkability, health data, and a dataset facilitating the linkage of cities from the walkability dataset to counties in the health dataset. Our analysis involved comparing the walkability scores across all states to people’s physical inactivity, as well as examining the correlation between walkability scores of counties in California to people’s physical inactivity and poor or fair health.

Based on our analysis, we found that poor or fair health is not sufficient to ascertain whether walkability directly leads to better health because external factors such as income levels, dietary habits, prevalence of chronic diseases, and accessibility to healthcare can impact the health status of individuals. When comparing physical inactivity to walkability, our results indicated that states with a higher walkability tend to have populations that are more physically active. We concluded that states with higher walkability had less physically inactive people, and those who are active typically experience better health than those who are not. However, as mentioned previously, it is important to recognize that health is influenced by multiple factors beyond physical activity alone, so we refrain from concluding that walkability itself directly causes improved health. Instead, we conclude that our findings suggest a correlation between physical activity and a state’s walkability score, implying that more walkable environments encourage healthier lifestyles.

Some of the main limitations of our project are the quality of the walkability data and the lack of a time frame specified for the data collected for the walkability index. As noted in our EDA process, there is a large variance for the walkability score of a single city, with scores ranging from [2,20] for a single city. With a walkability dataset with more specific columns relating the score to the time frame it was taken. This would allow us to understand the correlation on a more granular scale instead of reducing the dimensionality by taking the median or mean of the scores for a specific city. While this project does not have any immediate impact on society, if the question was explored more rigorously, the connection between walkability and health would be beneficial to improving the health of the individuals in areas with low walkability. If publicized, this work could be beneficial to establish more environmentally conscious cities and reduce carbon emissions while promoting walking and everyday activity for society.

Team Contributions

• Dashan McCain worked on the hypothesis, conclusion, one of EDA's graphs, and the project video.
• Luke Santa-Presca created the presentation and leaded the group on the project video.
• Leah Seseri contributed to the background and prior work, data overview, data cleaning, conclusion, and project video.
• Ya-Lun Wu worked on ethics & privacy, abstraction, and the project video.
• Kenny Zhang contributed to the prior work, data overview, data cleaning, EDA graphing, EDA analysis, and project video editing.