The recent Bernard Madoff Ponzi scheme debacle highlights that social networks can impose costs. In a series of articles, the New York Times has sketched how a group of insiders, mainly wealthy Jewish families from the New York region, sought to invest in Madoff's investment fund. Trust in Madoff as a man may have reduced effort in conducting due diligence to investigate whether his returns were too good to be true.

Social networks also offer large benefits. Successful executives and academics network constantly. How much of the returns to attending an Ivy League university are due to access to valuable social networks rather than what one learns from leading professors? An Ivy League graduate named Caroline Kennedy may soon be named the U.S Senator from New York. Her family and social connections appear to distinguish her from other ambitious professional politicians seeking the same senate seat such as Carolyn Maloney.

The fundamental challenge for empirical social scientists who want to study the causes and consequences of social networks is to identify who is the same network and to collect data on important outcomes that could be plausibly affected by participating in a network. For the last seven years, we have focused on the causes and consequences of social networks in a distinctive setting: the U.S Civil War. 2009 is the 200th anniversary of Abraham Lincoln's birth. What can we learn about social networks and social capital by studying the lives of enlisted men who fought for the Union Army?

In our new Princeton University Press book, Heroes and Cowards: The Social Face of War, we examine the war experience for Union Army soldiers. We weave a single narrative from the life histories of 41,000 Union Army soldiers, diaries and letters, and government documents. Our core questions are not those typically asked in a military history. When are men willing to sacrifice for the common good? What are the benefits to men of friendship? How do communities deal with betrayal? And what are the costs and benefits of being in a diverse community?

One summer we both read Robert Putnam's thought-provoking book Bowling Alone. We were fascinated by Putnam's account of the decline in American civic engagement over time. Putnam emphasized the growing popularity of television as a pivotal cause of the decline in social capital and community participation, but we wondered whether an unintended consequence of the rise of women working in the paid labor market was that PTAs and neighborhood associations lost their "volunteer army." We started to write a paper testing whether the rise in women's labor force participation explained the decline in residential community participation. To our surprise, we found little evidence supporting this claim. Instead, our analysis of long-run trends in volunteering, joining groups, and trust suggested that, all else equal, people who live in cities with more income inequality were less likely to be civically engaged. These results contributed to a growing literature in economics documenting the disturbing fact that people are less likely to be "good citizens" when they live in more diverse communities.

In the summer of 2001, we realized that the American Civil War provided the ideal "laboratory" for studying the costs and benefits of social networks. The setting was high stakes - roughly one out of every six Union Army soldiers died during the war. Unlike people in civilian life today, Union Army soldiers could not pick and choose their communities. For each of the 40,000 soldiers we observe key outcomes and choices. If a man deserts, if a man dies in a POW camp, if a man survives the war but chooses not to move back to his county of enlistment after the war, we observe each of these choices and outcomes. By studying how the probability of each of these outcomes varies as a function of individual solider attributes and the characteristics of the 100 men in his war community (his company), we quantify the role of social networks in a high stakes setting.

In our next post we will discuss our unique data set and why it is so difficult to create such a data set today.