In response to a recent essay about an apparentrelationship between urbanism and social alienation, Robert Steutevilleargues that the study in question - and its press - twists the facts.They say a lie can travel around the world in the time it takes for the truth to get its boots on. Richard Carson`s "Bowling Alone in Urbanistaville" recently published on the website ArchNewsNowrepeats false statements about a study that compares social activity inthe suburbs and the city. "The statistical revelation behind all ofthese findings is that for every 10% increase in density, there is a10% decrease in socialization. That`s a simple, one-to-one inverserelationship that everyone can understand," he writes.That`s not even close to being true — as Carson would know if he read the study. "Social Interaction and Urban Sprawl,"by Jan Brueckner and Ann Largey, finds that people talk to theirneighbors and hang out with friends more in cities than in suburbs on aper capita basis. They found no statistically significant difference inother observed social activity, such as the number of friends andconfidants, based on density.To fully explore how that simple truth got twisted 180 degrees, wehave to look more deeply at the study, its methods, its presentation,and how the authors publicized the findings. The `Unobservable` TendencyBrueckner and Largey use an instrumental-variables {IV} estimationto put numbers on what they call an "unobservable propensity," thetendency of gregarious people to locate in cities. IV estimation isused to correct for what the authors assume is the self-selection biasof people who want to live close to their neighbors.How do the authors know that something that they can`t detectactually exists and can be measured? They don`t. They assume that anormal person would seek to maximize lot size — that is, locate in asuburban place — all things being equal. Gregarious people, they say,like living closer to their neighbors and are willing to sacrifice lotsize in proportion to their desire to be social.In the authors` view, this would create higher social activity incities, and so the authors sought to take out the bias — that is,weight the results according to the tendency of outgoing people to livein denser environments. The entire credibility of the study dependsupon that assumption, which, as they say, is based on something that is"unobservable."Without the IV estimation, the study results are, as I have statedearlier, slightly in favor of cities. With the IV estimation, Bruecknerand Largey find an advantage for suburbs — but you have to go from very high density {11,591 people per square mile} to very low-density{779 people per square mile} to see any appreciable effect. This 93percent drop in density increases the chance of talking with neighborsby 5 percent, hanging out with friends by 7 percent, and having friendsover to your house by 13 percent. Why use such an extreme drop indensity? That`s because a less dramatic drop — say, 10 or 20 percent —yields numbers that are not statistically significant.Mixed-Up MessageSo much for the study and its methods, but its presentation andpublicity is even stranger. Brueckner and Largey bury thestraightforward, non-IV results of their study. They weren`t in thepress release or the introduction. Rather, you have to read deep into apaper full of barely readable jargon, comparing statements with tablesnot designed to enhance clarity. Here`s a sample of the authors` prose:"As can be seen, the density coefficients are insignificant in thenon-IV case in the NEISOC, CONFIDE, FRIENDS, FRNDHOM equations,becoming significantly negative when IV estimation is used. Bycontrast, the non-IV coefficients are significantly positive in theNEITALK and FRNDHNG equations, while IV estimation leads tocoefficients that are negative and insignificant." These acronyms referto social activity me
www.planetizen.com/node/34075