homework

POLI 100F Lecture 4: Social Networks and Health

Gregoire Phillips

University of California, San Diego

[email protected]

October 27, 2020

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 1 / 35

Overview

1 Announcements

2 Social Networks and Health

3 Recap

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 2 / 35

A few brief reminders

Participation Quiz #1 due tonight at midnight (can be found on Canvas under Assignments tab)

Project Pitch Memo due Friday!

Need to be submitted by Friday, Oct. 30th at 2PM PDT to be considered for full credit

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 3 / 35

A few brief reminders

Participation Quiz #1 due tonight at midnight (can be found on Canvas under Assignments tab)

Project Pitch Memo due Friday!

Need to be submitted by Friday, Oct. 30th at 2PM PDT to be considered for full credit

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 3 / 35

A few brief reminders

Participation Quiz #1 due tonight at midnight (can be found on Canvas under Assignments tab)

Project Pitch Memo due Friday!

Need to be submitted by Friday, Oct. 30th at 2PM PDT to be considered for full credit

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 3 / 35

Today’s Framing Questions

Our questions:

How do social networks affect our physical well-being?

How do social networks affect our psychological well-being?

Our goal today:

Explain how networks might contribute to the spread of behaviors that affect our health

Break down this process using the tools of the course

Discuss how social scientists test network theories using statistical methods (minimal math required)

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 4 / 35

Today’s Framing Questions

Our questions:

How do social networks affect our physical well-being?

How do social networks affect our psychological well-being?

Our goal today:

Explain how networks might contribute to the spread of behaviors that affect our health

Break down this process using the tools of the course

Discuss how social scientists test network theories using statistical methods (minimal math required)

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 4 / 35

Social Networks and Health Behaviors

Our day to day health is determined, in part, by the decisions we make about

What we consume

Whether we sleep

Whether or not we exercise

Many of these activities depend on our social connections to other people

Our friends and family influence our food habits

Our social and work lives influence our sleeping patters

Our friends, family, and work lives factor into our exercise choices

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 5 / 35

Social Networks and Health Behaviors

Our day to day health is determined, in part, by the decisions we make about

What we consume

Whether we sleep

Whether or not we exercise

Many of these activities depend on our social connections to other people

Our friends and family influence our food habits

Our social and work lives influence our sleeping patters

Our friends, family, and work lives factor into our exercise choices

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 5 / 35

Social Networks and Health Behaviors

Our day to day health is determined, in part, by the decisions we make about

What we consume

Whether we sleep

Whether or not we exercise

Many of these activities depend on our social connections to other people

Our friends and family influence our food habits

Our social and work lives influence our sleeping patters

Our friends, family, and work lives factor into our exercise choices

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 5 / 35

Social Networks and Health Behaviors

Our day to day health is determined, in part, by the decisions we make about

What we consume

Whether we sleep

Whether or not we exercise

Many of these activities depend on our social connections to other people

Our friends and family influence our food habits

Our social and work lives influence our sleeping patters

Our friends, family, and work lives factor into our exercise choices

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 5 / 35

Social Networks and Health Behaviors

Our day to day health is determined, in part, by the decisions we make about

What we consume

Whether we sleep

Whether or not we exercise

Many of these activities depend on our social connections to other people

Our friends and family influence our food habits

Our social and work lives influence our sleeping patters

Our friends, family, and work lives factor into our exercise choices

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 5 / 35

Social Networks and Health Behaviors

Our day to day health is determined, in part, by the decisions we make about

What we consume

Whether we sleep

Whether or not we exercise

Many of these activities depend on our social connections to other people

Our friends and family influence our food habits

Our social and work lives influence our sleeping patters

Our friends, family, and work lives factor into our exercise choices

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 5 / 35

Social Networks and Health Behaviors

Our day to day health is determined, in part, by the decisions we make about

What we consume

Whether we sleep

Whether or not we exercise

Many of these activities depend on our social connections to other people

Our friends and family influence our food habits

Our social and work lives influence our sleeping patters

Our friends, family, and work lives factor into our exercise choices

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 5 / 35

Social Networks and Health Behaviors

Our day to day health is determined, in part, by the decisions we make about

What we consume

Whether we sleep

Whether or not we exercise

Many of these activities depend on our social connections to other people

Our friends and family influence our food habits

Our social and work lives influence our sleeping patters

Our friends, family, and work lives factor into our exercise choices

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 5 / 35

Social Networks and Health Behaviors

We can think about health behaviors as spreading within our social networks

We and our friends, family, and acquaintances are the nodes

Ties describe friendship, familial bonds, or common interaction (in neighborhood or workplace, for example)

Ties can be unidirectional (I think of them as a friend, but they don’t think of me as one) Or ties can be bidirectional(we both think of each other as friends) This could be important, because it could determine the strength of the contagion within networks

The contagion is health behavior, or a behavior that has an impact on our health

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 6 / 35

Social Networks and Health Behaviors

We can think about health behaviors as spreading within our social networks

We and our friends, family, and acquaintances are the nodes

Ties describe friendship, familial bonds, or common interaction (in neighborhood or workplace, for example)

Ties can be unidirectional (I think of them as a friend, but they don’t think of me as one) Or ties can be bidirectional(we both think of each other as friends) This could be important, because it could determine the strength of the contagion within networks

The contagion is health behavior, or a behavior that has an impact on our health

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 6 / 35

Social Networks and Health Behaviors

We can think about health behaviors as spreading within our social networks

We and our friends, family, and acquaintances are the nodes

Ties describe friendship, familial bonds, or common interaction (in neighborhood or workplace, for example)

Ties can be unidirectional (I think of them as a friend, but they don’t think of me as one) Or ties can be bidirectional(we both think of each other as friends) This could be important, because it could determine the strength of the contagion within networks

The contagion is health behavior, or a behavior that has an impact on our health

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 6 / 35

Social Networks and Health Behaviors

We can think about health behaviors as spreading within our social networks

We and our friends, family, and acquaintances are the nodes

Ties describe friendship, familial bonds, or common interaction (in neighborhood or workplace, for example)

Ties can be unidirectional (I think of them as a friend, but they don’t think of me as one)

Or ties can be bidirectional(we both think of each other as friends) This could be important, because it could determine the strength of the contagion within networks

The contagion is health behavior, or a behavior that has an impact on our health

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 6 / 35

Social Networks and Health Behaviors

We can think about health behaviors as spreading within our social networks

We and our friends, family, and acquaintances are the nodes

Ties describe friendship, familial bonds, or common interaction (in neighborhood or workplace, for example)

Ties can be unidirectional (I think of them as a friend, but they don’t think of me as one) Or ties can be bidirectional(we both think of each other as friends)

This could be important, because it could determine the strength of the contagion within networks

The contagion is health behavior, or a behavior that has an impact on our health

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 6 / 35

Social Networks and Health Behaviors

We can think about health behaviors as spreading within our social networks

We and our friends, family, and acquaintances are the nodes

Ties describe friendship, familial bonds, or common interaction (in neighborhood or workplace, for example)

Ties can be unidirectional (I think of them as a friend, but they don’t think of me as one) Or ties can be bidirectional(we both think of each other as friends) This could be important, because it could determine the strength of the contagion within networks

The contagion is health behavior, or a behavior that has an impact on our health

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 6 / 35

Social Networks and Health Behaviors

We can think about health behaviors as spreading within our social networks

We and our friends, family, and acquaintances are the nodes

Ties describe friendship, familial bonds, or common interaction (in neighborhood or workplace, for example)

Ties can be unidirectional (I think of them as a friend, but they don’t think of me as one) Or ties can be bidirectional(we both think of each other as friends) This could be important, because it could determine the strength of the contagion within networks

The contagion is health behavior, or a behavior that has an impact on our health

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 6 / 35

Social Networks and Health Behaviors

We can think about health behaviors as spreading within our social networks

We and our friends, family, and acquaintances are the nodes

Ties describe friendship, familial bonds, or common interaction (in neighborhood or workplace, for example)

Ties can be unidirectional (I think of them as a friend, but they don’t think of me as one) Or ties can be bidirectional(we both think of each other as friends) This could be important, because it could determine the strength of the contagion within networks

The contagion is health behavior, or a behavior that has an impact on our health

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 6 / 35

Modeling Health Behavior Transmission

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 7 / 35

Modeling Health Behavior Transmission

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 8 / 35

Modeling Health Behavior Transmission

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 9 / 35

Modeling Health Behavior Transmission

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 10 / 35

Modeling Health Behavior Transmission

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 11 / 35

Social Networks and Health Behaviors

Case example: social networks and smoking

Smoking is a textbook example of the contagion of health behaviors in networks

James Fowler and Nicholas Christakis conducted a study on this behavior to determine how contagious this behavior is

Took sample from a famous biomedical study, the Framingham Heart Study, that looked at the evolution of health behaviors among the same group of people in a large social network over 20 years (longitudinal study)

Look at how people picking up and quitting the habit of smoking changes within the network of 5,124 people

Isolate the effect of other people’s behavior (alter behavior) on each ego node.

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 12 / 35

Social Networks and Health Behaviors

Case example: social networks and smoking

Smoking is a textbook example of the contagion of health behaviors in networks

James Fowler and Nicholas Christakis conducted a study on this behavior to determine how contagious this behavior is

Took sample from a famous biomedical study, the Framingham Heart Study, that looked at the evolution of health behaviors among the same group of people in a large social network over 20 years (longitudinal study)

Look at how people picking up and quitting the habit of smoking changes within the network of 5,124 people

Isolate the effect of other people’s behavior (alter behavior) on each ego node.

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 12 / 35

Social Networks and Health Behaviors

Case example: social networks and smoking

Smoking is a textbook example of the contagion of health behaviors in networks

James Fowler and Nicholas Christakis conducted a study on this behavior to determine how contagious this behavior is

Took sample from a famous biomedical study, the Framingham Heart Study, that looked at the evolution of health behaviors among the same group of people in a large social network over 20 years (longitudinal study)

Look at how people picking up and quitting the habit of smoking changes within the network of 5,124 people

Isolate the effect of other people’s behavior (alter behavior) on each ego node.

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 12 / 35

Social Networks and Health Behaviors

Case example: social networks and smoking

Smoking is a textbook example of the contagion of health behaviors in networks

James Fowler and Nicholas Christakis conducted a study on this behavior to determine how contagious this behavior is

Took sample from a famous biomedical study, the Framingham Heart Study, that looked at the evolution of health behaviors among the same group of people in a large social network over 20 years (longitudinal study)

Look at how people picking up and quitting the habit of smoking changes within the network of 5,124 people

Isolate the effect of other people’s behavior (alter behavior) on each ego node.

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 12 / 35

Social Networks and Health Behaviors

Case example: social networks and smoking

Smoking is a textbook example of the contagion of health behaviors in networks

James Fowler and Nicholas Christakis conducted a study on this behavior to determine how contagious this behavior is

Took sample from a famous biomedical study, the Framingham Heart Study, that looked at the evolution of health behaviors among the same group of people in a large social network over 20 years (longitudinal study)

Look at how people picking up and quitting the habit of smoking changes within the network of 5,124 people

Isolate the effect of other people’s behavior (alter behavior) on each ego node.

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 12 / 35

Social Networks and Health Behaviors

Case example: social networks and smoking

Smoking is a textbook example of the contagion of health behaviors in networks

James Fowler and Nicholas Christakis conducted a study on this behavior to determine how contagious this behavior is

Took sample from a famous biomedical study, the Framingham Heart Study, that looked at the evolution of health behaviors among the same group of people in a large social network over 20 years (longitudinal study)

Look at how people picking up and quitting the habit of smoking changes within the network of 5,124 people

Isolate the effect of other people’s behavior (alter behavior) on each ego node.

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 12 / 35

Framingham Heart Study

Classic example of a really, really good data source for network analysis

Survey of 5,124 people – called egos in study of networks

longitudinal study – the same people were tracked over time, and new ties were accounted for in later waves of the survey (eight in total so far)

relatively balanced sample of male and female identifying individuals

Sampled through common social connections to ensure capture of network – spouses, friends, family members

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 13 / 35

Framingham Heart Study

Classic example of a really, really good data source for network analysis

Survey of 5,124 people – called egos in study of networks

longitudinal study – the same people were tracked over time, and new ties were accounted for in later waves of the survey (eight in total so far)

relatively balanced sample of male and female identifying individuals

Sampled through common social connections to ensure capture of network – spouses, friends, family members

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 13 / 35

Social Networks and Smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 14 / 35

Social Networks and Smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 15 / 35

Social Networks and Smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 16 / 35

Social Networks and Smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 17 / 35

Social Networks and Smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 18 / 35

Social Networks and Smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 19 / 35

Social Networks and Smoking

Smoking and non-smoking friends influence behavior significantly within three degrees of separation

Surprisingly, geography plays less of a role than we would expect

Smokers and non-smokers tend to cluster within large social networks

As smoking becomes less socially popular, non-smokers become more central within this network, speeding up its decline

Friends (particularly better educated ones), mutual friends, spouses, siblings, and coworkers all had significant impacts on decreasing risk of smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 20 / 35

Social Networks and Smoking

Smoking and non-smoking friends influence behavior significantly within three degrees of separation

Surprisingly, geography plays less of a role than we would expect

Smokers and non-smokers tend to cluster within large social networks

As smoking becomes less socially popular, non-smokers become more central within this network, speeding up its decline

Friends (particularly better educated ones), mutual friends, spouses, siblings, and coworkers all had significant impacts on decreasing risk of smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 20 / 35

Social Networks and Smoking

Smoking and non-smoking friends influence behavior significantly within three degrees of separation

Surprisingly, geography plays less of a role than we would expect

Smokers and non-smokers tend to cluster within large social networks

As smoking becomes less socially popular, non-smokers become more central within this network, speeding up its decline

Friends (particularly better educated ones), mutual friends, spouses, siblings, and coworkers all had significant impacts on decreasing risk of smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 20 / 35

Social Networks and Smoking

Smoking and non-smoking friends influence behavior significantly within three degrees of separation

Surprisingly, geography plays less of a role than we would expect

Smokers and non-smokers tend to cluster within large social networks

As smoking becomes less socially popular, non-smokers become more central within this network, speeding up its decline

Friends (particularly better educated ones), mutual friends, spouses, siblings, and coworkers all had significant impacts on decreasing risk of smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 20 / 35

Social Networks and Smoking

Smoking and non-smoking friends influence behavior significantly within three degrees of separation

Surprisingly, geography plays less of a role than we would expect

Smokers and non-smokers tend to cluster within large social networks

As smoking becomes less socially popular, non-smokers become more central within this network, speeding up its decline

Friends (particularly better educated ones), mutual friends, spouses, siblings, and coworkers all had significant impacts on decreasing risk of smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 20 / 35

Social Networks and Smoking

Smoking and non-smoking friends influence behavior significantly within three degrees of separation

Surprisingly, geography plays less of a role than we would expect

Smokers and non-smokers tend to cluster within large social networks

As smoking becomes less socially popular, non-smokers become more central within this network, speeding up its decline

Friends (particularly better educated ones), mutual friends, spouses, siblings, and coworkers all had significant impacts on decreasing risk of smoking

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 20 / 35

Social Networks and Drinking

This effect is remarkably similar within drinking behavior

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 21 / 35

Social Networks and Drinking

This effect is remarkably similar within drinking behavior

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 22 / 35

Social Networks and Drinking

This effect is remarkably similar within drinking behavior

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 23 / 35

Social Networks and Drinking

This effect is remarkably similar within drinking behavior

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 24 / 35

Social Networks and Happiness

Case example: Social networks and happiness

Emotions are notoriously “contagious” – but can they actually spread through a social network?

Claim: we are bioloigcally hardwired to outwardly mimic others

Claim 2: We benefit, evolutionarily, from adopting the inward state of others

Christakis and Fowler also tested this

Used the same study – Framingham Heart Study – to test network effects of happiness

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 25 / 35

Social Networks and Happiness

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 26 / 35

Social Networks and Happiness

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 27 / 35

Social Networks and Happiness

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 28 / 35

Social Networks and Happiness

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 29 / 35

Social Networks and Happiness

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 30 / 35

Social Networks and Happiness

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 31 / 35

Social Networks and Happiness

Hold on, though. Why are we seeing clustering? Is it because the behavior is contagious?

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 32 / 35

Social Networks and Happiness

Fit an econometric model that takes into account individual, time, and geographic fixed effects to control for ommitted variables

Use model and data features itself to assess selection.

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 33 / 35

Social Networks and Happiness

Results strongly suggest a contagion effect

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 34 / 35

What we learned

Social networks can impact our health through the proliferation of health behaviors

We can leverage simple survey methods – selecting a sample of people and building out their social networks – to measure the effect of behavior over time

We can use econometric tools to test the effect of the behaviors of others within a network and measure the strength of this effect over time, across distances, and across ties

NEXT TIME: Social Networks and Social Life!

Gregoire Phillips (UCSD) Social Data Analysis October 27, 2020 35 / 35