Judith Hellerstein. Social Capital

A lot of recent evidence that networks that match workers to employers are productive.

Neighborhood-based labor market networks are important too.

·       where workers work, (bayer et al 2008; HMN 2011)

·       tenure and earnings (Schmutte 2015, HKN 2014)

·       re-employment, earnings tenure (HKN 2016)

Moving to Opportunity (MTO).

What makes some neighborhoods more “networked” at work than others?

Is there an important role for social capital? What makes more neighborhoods more networked than others?

Social Capital: “The networks of relationships among people who live and work in a particular society, enabling that society to function effectively (OED definition).” Effectively à positive labor market outcomes.

Chetty et al (2014) who examine the relationship between social capital measures and intergenerational mobility at CZ level.

Data: “Who did you seek out when seeking a job?” in a survey.

Avoid: ex ante selection. Data mininign for significant predictors easiest to rationalize ex post as social capital measures

Combine data from: government administrative data (LEHD). Public Use data (ACS, Common Core , HEDA); Private administrative data (NETS). à better coverage of non-profits than LBD.

They use machine learning to find the variables that matter. That is, they let the data tell them what matters in social capital measure.

Networks lower information ost about existence of vacancies to unempllyed job searchers (Calvo armengol and Jackson 2007)

Networks lower information costs to employers about productivity of potential hires. (Montgomery 1991)

What is inferred is Networks are productive. They aren’t always. Their evidence suggests they are productive. We are not thinking that it is zero sum. One benefits at the cost of another. That’s a major criticism of social capital, that it excludes people. Mostly those who benefit are the low skilled people who find jobs in the neighborhood.

Four categories of social capital

-          Demographic homogeneity

-          School inputs

-          Civic participation/homogeneity (Guiso et al 2004)

-          Non-profit activity

o   Civic institutions - libraries, community centers (Coleman 1998)

o   Religious organizations (Putnam 2000)

o   Clubs (Putnam 2000)

Bowling leagues have declined. Now we bowl alone.

Individual’s network solation index.

Fraction of workers that you work with who come from your neighborhood census tract.

Evidence shows that networks are race based.

Machine learning

Estimation from objective function (Belloni et al 2014).

OLS and LASSO results.

What matters?

Amongst number of districts, student/teacher ratio, free/reduced price lunch share, majority vote share, democratic vote share, voter turnout…

What mattered was student/teacher ratio and democratic vote share.

Signs are mostly consistent with expectations

-          More districts, more networked

-          Smaller classes, more networked

-          Higher democratic vote share: less networked.