Social withdrawal in the US: objective social isolation, subjective loneliness, or chosen solitude?

[chinaexpert1]

Investigate America’s Shrinking Social World

Overview

Americans are spending more time at home and less time socializing in person than they did two decades ago. In 2025, only 30% of Americans engaged in socializing and communicating on an average day, compared with 38% in 2015; average time spent in these activities declined from 41 to 35 minutes. Watching television occupied approximately half of all leisure time.

Longer-term American Time Use Survey analysis indicates that Americans spend about 99 additional minutes at home each day compared with 2003. Among people ages 15–24, the increase is approximately 124 minutes.

This retreat may reflect several overlapping changes:

• Remote work;
• Digital entertainment and social media;
• Home delivery and online shopping;
• Declining civic and religious participation;
• Loss of affordable gathering places;
• Long commuting times;
• Housing and transportation design;
• Financial pressure;
• Aging and living alone;
• Pandemic-era habits that persisted.

The trend is not automatically harmful. Working from home may improve accessibility, family life, and time autonomy. Solitude may be restorative. The policy concern arises when reduced in-person interaction produces unwanted loneliness, weak social support, declining civic participation, poorer health, or reduced resilience.

The central research question is:

Are Americans becoming measurably more socially isolated, which structural conditions explain the change, and which community interventions successfully increase meaningful social connection?

The recommended primary units of analysis are:

• Individual-year for survey analysis;
• County- or census-tract-year for geographic analysis;
• Neighborhood or intervention site for program evaluation.

---

Action Items

Initial Evidence

Current evidence justifies a national investigation:

• Americans are less likely to socialize in person on an average day than they were ten years ago.
• Americans spend substantially more time at home than in 2003, with especially large increases among young people.
• The 2024 Household Pulse Survey found that loneliness is strongly associated with weak social, economic, and community support, although most respondents reported feeling lonely rarely or never.
• The Census Household Pulse Survey added questions about loneliness, social support, and social isolation beginning in January 2024.
• CDC PLACES now includes local estimates for loneliness or social isolation and other health-related social needs in 39 states and the District of Columbia.
• The U.S. Surgeon General has identified social isolation as a public-health risk and reports that isolation is associated with a 29% higher risk of premature mortality.
• Recent periodical commentary links reduced social activity to disappearing “third places,” expensive outings, digital substitution, neighborhood design, and the increasing convenience of remaining at home.
• Community-based programs are already attempting to rebuild local connection through public spaces, shared activities, mutual aid, food programs, and neighborhood organizations.

These observations establish a change in behavior and a public-health concern, but they do not prove that phones, remote work, urban design, or any other single factor is responsible.

---

Hypotheses

H1 — Digital substitution is replacing in-person interaction

Streaming, gaming, social media, online shopping, delivery services, and remote communication may replace activities that once required physical presence.

Evidence to seek:

• Screen-based leisure increases as in-person social time declines;
• Digital adoption precedes changes in social behavior;
• Effects are largest among adolescents and young adults;
• Online interaction supplements connection for some groups but replaces it for others;
• Increased digital time predicts loneliness only when it displaces meaningful relationships.

H2 — Remote work reduces weak social ties

Working from home may preserve close relationships while reducing routine contact with coworkers, commuters, shopkeepers, and other acquaintances.

Evidence to seek:

• Fully remote workers report fewer daily in-person contacts;
• Hybrid workers differ from fully remote and fully onsite workers;
• Remote-work effects depend on household composition;
• Social outcomes improve when workers have structured coworking or community activity;
• People with disabilities or caregiving responsibilities experience different effects.

H3 — Loss of affordable third places reduces social participation

Libraries, parks, recreation centers, cafés, diners, bars, churches, clubs, malls, and community organizations provide low-barrier settings for repeated interaction.

Evidence to seek:

• Communities losing third places experience larger declines in social activity;
• Library, park, or recreation access predicts lower loneliness;
• Business closures affect young, low-income, and older residents differently;
• New or renovated public spaces increase visits and repeated interaction;
• Affordability matters more than the simple number of venues.

H4 — Car-dependent neighborhood design suppresses spontaneous interaction

Low-density development, long travel distances, limited sidewalks, and weak transit may reduce opportunities for casual social contact.

Evidence to seek:

• Walkability predicts more nonwork trips and local interaction;
• Long commute times predict lower civic and social participation;
• Transit-accessible neighborhoods show different patterns after income controls;
• Mixed-use neighborhoods retain more social activity;
• Effects vary by age, disability, and household type.

H5 — Financial pressure makes social participation unaffordable

Dining, entertainment, childcare, transportation, and event costs may discourage participation, especially among younger and lower-income adults.

Evidence to seek:

• Social activity declines as local leisure costs rise;
• Income and debt predict social participation after time constraints are controlled;
• Free public amenities reduce income-related social gaps;
• Working multiple jobs reduces both available time and social activity;
• Cost predicts staying home more strongly than stated preference.

H6 — Declining civic institutions reduce repeated group participation

Lower participation in religious organizations, unions, clubs, neighborhood groups, volunteer organizations, and local associations may reduce durable social networks.

Evidence to seek:

• Declining membership predicts weaker perceived support;
• Volunteering and group participation protect against loneliness;
• Counties with stronger civic institutions show greater resilience after disasters;
• Institutional closures produce measurable local effects;
• Repeated participation matters more than one-time events.

H7 — Living alone and delayed household formation increase risk for some groups

Growth in one-person households may increase isolation, although living alone should not be treated as equivalent to loneliness.

Evidence to seek:

• Living alone predicts loneliness only under certain age, income, or health conditions;
• Neighborhood connection moderates the effect;
• Older adults and young adults show different patterns;
• Shared housing reduces isolation only when relationships are supportive;
• Housing instability disrupts social networks.

H8 — Social isolation is partly a transportation-access problem

People may have relationships and desired activities but lack safe, affordable, or reliable transportation.

Evidence to seek:

• Transportation barriers predict loneliness after demographic controls;
• Transit reductions precede declining participation;
• Paratransit access improves connection among older and disabled adults;
• Car ownership has nonlinear effects in different neighborhood types;
• Travel time to social infrastructure predicts participation.

H9 — Social isolation contributes to poor health and reduced resilience

Weak social connection may affect depression, cardiovascular disease, cognitive decline, mortality, emergency preparedness, and recovery from illness.

Evidence to seek:

• Isolation precedes deterioration rather than merely appearing after illness;
• Effects remain after baseline health controls;
• Stronger local networks predict better disaster recovery;
• Social-support interventions improve measurable health outcomes;
• Effects differ between loneliness and objective isolation.

H10 — The apparent national trend consists of different population-specific problems

Young adults, parents, older adults, rural residents, remote workers, people with disabilities, and recent migrants may experience different forms of disconnection.

Evidence to seek:

• Distinct social-isolation clusters emerge;
• Causes and protective factors differ across groups;
• National averages conceal areas with strong connection;
• The same intervention has heterogeneous effects;
• Some groups gain connection through technologies that isolate others.

---

Research Plan

1. Define separate outcomes

Measure at least four dimensions:

Objective isolation

• Number of in-person contacts;
• Frequency of social activity;
• Time spent alone;
• Household composition;
• Group participation.

Subjective loneliness

• Frequency of feeling lonely;
• Satisfaction with relationships;
• Perceived lack of companionship.

Social support

• Availability of emotional or practical help;
• Frequency of communication;
• Number of trusted contacts.

Community connection

• Volunteering;
• Civic participation;
• Neighborhood trust;
• Institutional membership;
• Cross-class or cross-group relationships.

Do not combine these into one score until each component has been analyzed independently.

2. Establish the national timeline

Use American Time Use Survey microdata from 2003 onward to measure:

• Time at home;
• Socializing and communicating;
• Entertainment;
• religious and civic activity;
• volunteering;
• sports;
• travel;
• work at home;
• time alone where available.

Separate:

• 2003–2012;
• 2013–2019;
• 2020–2022;
• 2023 onward.

Use segmented regression or change-point detection to determine whether the decline was gradual or accelerated around specific years.

3. Build an individual-level analytical dataset

Join or compare compatible measures from:

• American Time Use Survey;
• Household Pulse Survey;
• Behavioral Risk Factor Surveillance System;
• General Social Survey;
• National Health Interview Survey;
• Current Population Survey volunteering and civic supplements.

Variables should include:

• Age;
• sex;
• income;
• education;
• employment;
• remote-work status;
• marital and household status;
• disability;
• urbanicity;
• transportation;
• physical and mental health;
• social activity;
• loneliness;
• social support.

Use survey weights and survey-design corrections.

4. Build a county- or tract-level social infrastructure dataset

Measure access to:

• Libraries;
• parks;
• recreation centers;
• houses of worship;
• cafés and restaurants;
• bars;
• community centers;
• senior centers;
• transit;
• sidewalks and bike networks;
• mixed-use commercial areas.

Calculate:

Where (k) represents a type of social infrastructure and the distance-decay function gives more weight to nearby destinations.

Do not assume every venue produces social connection. Distinguish passive access from actual usage.

5. Measure third-place decline

Track:

• Establishment openings and closures;
• Library branches and visits;
• recreation facilities;
• religious congregations;
• local retail;
• bars and restaurants;
• civic associations;
• mall and community-space redevelopment.

Test whether changes in third-place density predict changes in:

• Time outside the home;
• loneliness;
• volunteering;
• neighborhood trust;
• pedestrian activity.

6. Evaluate remote-work effects

Compare:

• Fully remote;
• hybrid;
• fully onsite;
• unemployed or retired populations.

Estimate:

Potential natural experiments include employer return-to-office changes or variation in occupational telework feasibility.

Avoid assuming that office presence automatically creates meaningful connection.

7. Test neighborhood-design hypotheses

Combine:

• Walkability;
• land-use mix;
• population density;
• street connectivity;
• transit service;
• commute time;
• traffic danger;
• access to parks and libraries.

Use matched neighborhood comparisons or changes following:

• New transit stations;
• pedestrian improvements;
• library openings;
• park renovations;
• removal of parking requirements;
• mixed-use redevelopment.

Measure actual social and travel behavior before and after the intervention.

8. Evaluate community interventions

Candidate interventions include:

• Library programming;
• public-space improvements;
• recurring neighborhood events;
• community gardens;
• intergenerational programs;
• sports and arts groups;
• social prescribing;
• volunteer programs;
• coworking spaces;
• free or reduced-fare transit.

For each program, record:

• Eligibility;
• frequency;
• attendance;
• retention;
• cost;
• participant characteristics;
• comparison group;
• loneliness and support before and after;
• longer-term participation.

Randomized encouragement designs may be more feasible than forcing program assignment.

9. Examine health consequences

Model whether isolation predicts subsequent:

• Depression;
• poor self-rated health;
• chronic disease;
• hospital utilization;
• cognitive decline;
• mortality.

Use longitudinal data where possible. Cross-sectional correlations cannot establish direction because poor health can cause isolation.

10. Identify positive-deviant communities

Find communities with:

• High social connection despite low income;
• Low loneliness despite many one-person households;
• Strong civic participation despite low density;
• Recovery in social activity after the pandemic;
• High usage of public social infrastructure.

Compare them with demographically similar areas to identify potentially transferable conditions.

11. Build a Social Connection Dashboard

Recommended measures:

• In-person social minutes per day;
• Percentage socializing on an average day;
• Loneliness prevalence;
• Social-support prevalence;
• Time spent at home;
• Volunteering;
• Library visits;
• Third places per 10,000 residents;
• Transit and walkability access;
• Remote-work prevalence;
• One-person households;
• Community-group participation.

Show uncertainty and avoid ranking small areas with unstable estimates.

12. Produce the Final PowerPoint

Recommended structure:

1. America’s retreat into the home;
2. Isolation, loneliness, and solitude;
3. National time-use trend;
4. Age and demographic differences;
5. Remote work and digital substitution;
6. Housing, transportation, and neighborhood design;
7. Decline of third places;
8. Financial barriers to participation;
9. Health and resilience consequences;
10. Positive-deviant communities;
11. Interventions with measurable evidence;
12. Policy options and evaluation metrics.

The presentation should distinguish:

• Chosen solitude from unwanted loneliness;
• Online communication from meaningful support;
• Living alone from isolation;
• Correlation from causal evidence;
• Access to a place from actual participation;
• One-time attendance from durable connection.

---

Resources

Data, APIs, and Where to Get Data

American Time Use Survey

The Bureau of Labor Statistics provides annual microdata, activity codes, respondent characteristics, and survey weights dating to 2003. The latest 2025 release reports declining participation and time in socializing compared with 2015.

Use for:

• Time spent at home;
• socializing;
• volunteering;
• religious activity;
• leisure;
• travel;
• remote work;
• demographic comparisons.

ATUS is the principal source for measuring long-run behavioral change.

Census Household Pulse Survey API and Public-Use Files

The Household Pulse Survey includes recent questions about loneliness, social support, and social isolation. Its high-frequency estimates are accessible through the Census Data API, and public-use files are available for individual-level analysis.

Use for:

• Loneliness;
• social support;
• demographic disparities;
• employment and remote work;
• housing and financial stress;
• state and metropolitan comparisons.

Caution: the Census Bureau classifies this as an experimental data product, and the cross-sectional collection ended in September 2024.

CDC PLACES and Socrata API

CDC PLACES publishes local model-based estimates at county, place, census-tract, and ZIP Code Tabulation Area levels. The 2025 release includes health-related social-needs measures, including loneliness or social isolation in participating states.

Use for:

• Local loneliness estimates;
• health outcomes;
• mental and physical health;
• transportation barriers;
• lack of social support;
• disability and chronic disease.

Caution: PLACES values are modeled estimates, not direct local survey counts.

Behavioral Risk Factor Surveillance System

Use for:

• Social and emotional support;
• health status;
• depression;
• disability;
• chronic conditions;
• state-level subgroup analysis.

BRFSS microdata require survey weighting and careful treatment of changing modules.

National Health Interview Survey

Use for:

• Social connection;
• health;
• disability;
• mental distress;
• household characteristics;
• healthcare access.

Opportunity Insights Social Capital Atlas

Opportunity Insights provides downloadable social-capital measures by ZIP code and county. Measures include economic connectedness, social cohesion, and civic engagement derived partly from large-scale friendship-network data.

Use for:

• Cross-class relationships;
• volunteering;
• social cohesion;
• geographic comparison;
• connections with economic mobility.

Caution: social-media users are not a perfect representation of the entire population.

Institute of Museum and Library Services Public Libraries Survey

The Public Libraries Survey covers approximately 9,000 public-library systems and about 17,000 outlets nationwide.

Use for:

• Library branches;
• visits;
• programs;
• attendance;
• staffing;
• operating hours;
• community-service availability.

Libraries are particularly useful because they are free, geographically distributed third places.

National Household Travel Survey

Use for:

• Trip purpose;
• travel time;
• mode;
• household vehicles;
• nonwork travel;
• demographic travel differences.

This can help distinguish lack of desired social activity from lack of transportation access.

Census American Community Survey API and PUMS

Use for:

• One-person households;
• household composition;
• remote work;
• commuting;
• income;
• age;
• disability;
• housing costs;
• migration;
• neighborhood demographics.

Census LEHD and LODES

Use for:

• Home-to-work flows;
• job accessibility;
• commute geography;
• employment centers;
• neighborhood-worker connections.

OpenStreetMap

Use for geospatial inventories of:

• Parks;
• cafés;
• libraries;
• community centers;
• religious facilities;
• pedestrian infrastructure;
• transit stops;
• recreation facilities.

Validate local completeness before comparing places.

Census County Business Patterns

Use for:

• Restaurants;
• bars;
• entertainment;
• recreation;
• religious and civic organizations;
• establishment density;
• openings and closures where linked to other business datasets.

Census Business Dynamics Statistics

Use for:

• Establishment entry and exit;
• local business turnover;
• industry-level employment change.

IMLS Public Needs for Library and Museum Services Survey

This survey measures awareness, visitation, attitudes, and public demand for library and museum services.

General Social Survey

Use for long-term trends in:

• Trust;
• close relationships;
• organizational membership;
• religion;
• civic attitudes;
• social participation.

The General Social Survey is not annual, so it should complement rather than replace ATUS.

Current Population Survey Civic Engagement and Volunteering Supplements

Use for:

• Volunteering;
• group participation;
• civic activity;
• neighborhood engagement;
• demographic differences.

Local Administrative Data

Potential sources include:

• Library card and program attendance;
• recreation registration;
• park counters;
• transit ridership;
• community-center use;
• event permits;
• pedestrian counts;
• nonprofit membership;
• 311 requests;
• local business licenses.

These data may provide stronger intervention measures than national surveys but require local privacy review.

Foundational Resources

• Bureau of Labor Statistics, American Time Use Survey—2025 Results.
• Washington Post, America Is Becoming a Nation of Homebodies.
• The Atlantic, The Anti-Social Century.
• TIME, Why a “Third Life” Is the Answer to America’s Loneliness Epidemic.
• U.S. Surgeon General, Social Connection.
• Opportunity Insights, Social Capital I: Measurement and Associations with Economic Mobility.
• CDC, PLACES health-related social-needs definitions and methodology.

Central Scientific Caution

Loneliness is subjective, socially influenced, and sensitive to question wording. Objective isolation, time at home, living alone, and loneliness are related but noninterchangeable.

The strongest investigation will triangulate:

1. Time-use behavior;
2. Self-reported loneliness;
3. Social support;
4. Household structure;
5. Physical mobility;
6. Civic participation;
7. Social-infrastructure access;
8. Actual facility use;
9. Health outcomes.

The best policy target may not be “reduce time at home.” It may be to increase voluntary, meaningful, repeated connection while preserving the genuine benefits of solitude, remote work, disability access, and home-centered family life.

• If this issue requires access to 311 data, please answer the following questions:
  • Do you need a one-time or ongoing dump of the data?
  • Do you need subset of data (i.e. certain years) or the entire data set (approx. 4 million rows or 11 GB)?
    • If a subset is needed, please define subset characteristics (i.e. date range, etc.)
  • Do you need online access via an API or a download of data?