NOTES

CHAPTER ONE

1. Elsewhere USA explores the seeming fixation of Americans on connecting to everywhere but where they actually are (Conley 2009).

2. Giddens 1990, 19; 1991, 146. Giddens’s (1991) treatise on modernity has been cited over 14,000 times in the scholarly literature, perhaps an unparalleled record. For an influential statement on placelessness, cited less but still a citation classic at over two thousand, see Relph 1976. For an argument that the “death of distance” is “changing the world,” see Cairncross 1997.

3. Friedman 2005.

4. The seminal empirical adjudication of personal networks in the modern era is Claude Fischer (1982), who showed that urbanism’s effects are specified; estrangement occurs in the public sphere—less helpfulness and more conflict—but not in the private sphere of personal relationships and personal wellbeing. See also Tilly 1973, 211.

5. Wellman 1979.

6. Brint 2001.

7. Louis Wirth (1938) was thus one of the first to link diversity with decline. I return to this point in later chapters when delving empirically into the puzzle of neighborhood heterogeneity. Interestingly, as Everett Hughes pointed out in personal communication to James F. Short Jr.: “Louis used to say all those things about how the city is impersonal while living with a whole clan of kin and friends on a very personal basis.” Apparently Wirth’s theoretical angst about the decline of community was not realized in the social world he inhabited. One wonders if this is true of contemporary doomsayers, including those chatting or texting away.

8. Wellman 1979. See also Fischer 1982, chapter 1. The history of social disorganization theory is addressed in chapter 4; see also Kornhauser 1978.

9. Nisbet 1953.

10. See especially Putnam 2000, 2007.

11. Coleman 1988. The application to urban sociology and community has been widespread. A recent example is Saegert, Thompson, and Warren 2002.

12. Smelser and Alexander 1999; Skocpol 2004.

13. McAdam, McCarthy, and Zald 1996.

14. See the Institute for Communitarian Policy Studies, (http://www.gwu.edu/^~ccps/index.html), and its former journal, The Responsive Community.

15. McPherson, Smith-Lovin, and Brashears 2006.

16. Putnam 2007.

17. See Simmel, “The Metropolis and Mental Life” (1903), reprinted in Levine 1971.

18. Selznick 1992; Etzioni 1997.

19. Mailer 1969, opening line, Part II

20. Abu-Lughod 1999.

21. The “Los Angeles School” of urban sociology (e.g., Dear 2002) might be thought to be the leading proponent of this view, but I refer broadly to current thinking across the social sciences, including sociology, political science, epidemiology, and perhaps especially economics. Indeed, as will be seen in examples throughout the book (especially chapters 11–12), the idea that neighborhoods “do not matter” in an individually rationalistic and placeless world is commonplace despite the long existence of research on neighborhood effects (chapter 2). The LA School of geography seeks to answer a different but legitimate class of questions (Sampson 2002a; Abbott 2002).

22. Zorbaugh 1929, 13.

23. The tour begins here in Chicago’s Near North Side (later, compare fig. 1.2): http://www.mapquest.com/maps?city=Chicago&state=IL&address=N+Michigan+Ave+%26+E+Pearson+St&zipcode=60611&country=US&latitude=41.89757&longitude=-87.62411&geocode=INTERSECTION.

24. Wilson and Kelling 1982.

25. During a street observation on another walk in 2006 (March 10), I witnessed a protest march replete with drummers, chanting, and dissemination of a statement from the Dalai Lama. Leading the march were monks who sought to draw attention to Tibetan dissidents on the occasion of the forty-seventh anniversary of the Tibetan National Uprising Day. The nature and location of this event corresponds to a theory of collective civic engagement pursued later in the book.

26. Emporis Buildings 2007.

27. Algren 1951.

28. One might be forgiven for noting similarities between the man, the building, and the granddaddy of Chicago politics, Richard J. Daley. “He was raucous, sentimental, hot-tempered, practical, simple, devious, big, and powerful. This is, after all, Chicago” (Royko 1971).

29. Gilfoyle 2006.

30. Bridgeport, an iconic working-class white neighborhood that produced generations of political heavyweights, sits south and west of the Dan Ryan, within earshot of the Chicago White Sox’s Comiskey Park. The Daleys’ move away from Bridgeport to a nearly half-million-dollar townhouse in Central Station Townhouse (in 1994 dollars) was the subject of much consternation among locals and widely reported in the local media. It was not the price so much as the identity of location—the beers and brats of Richard M.’s father in Bridgeport were given a symbolic punch. Daley #2 has been reported as intending to move up even more, quite literally, to a high-rise perch across from Millennium Park, smack in the heart of the Loop. Rising over a former Dunkin’ Donuts near the corner of Wabash and Randolph, the Heritage at Millennium Park commands stunning views of the park and Lake Michigan to the east.

31. Nels Anderson’s On Hobos and Homelessness is widely considered the major “Chicago School” take on the homeless—see Anderson 1998. He studied the West Madison Street Area, but the Near South was another favorite haunt of hobos in the early years. At the current intersection stands a Latin American and Caribbean fusion restaurant, where I dined twice in 2007 with many well-heeled and fashionable people out on the town.

32. I return later to gentrification, but it is not clear the term fits the Near South. Gentrification is commonly thought to mean areas where the lower-income residents are pushed out. Many areas here had no permanent residents. The South Loop has seen a net loss of thousands of SRO units since 1990.

33. For a sociological history of the Robert Taylor Homes, see Venkatesh 2000.

34. Chicago Housing Authority 2007.

35. Whether crime, school dropout, mortality, or disease, the Wentworth district long stood out as a leader in what William Julius Wilson (1987) called “social dislocation.”

36. Field observations, 2007.

37. For a description of the changes in the North Kenwood area, especially the “gentrification” conflict between lower- and upper-class African Americans, see Pattillo 2007.

38. Typical is the 8000 South Avalon block club, welcoming visitors but “no ball playing, car washing, speeding, loud music, loitering or littering.” Nearly every block declares the same.

39. Zorbaugh 1929, 14.

40. In introducing the bird’s-eye view, I map the community areas in Chicago. Lake Michigan is the defining natural feature of the city’s border to the east—the city boasts some twenty-five miles of lakefront property. A less beautiful feature of Chicago is O’Hare International Airport extending westward from the city in the northwest corner, a place likely to have been experienced by every reader. Community areas are defined in more detail in chapter 2, and individual communities important to the story of this book will be mapped and described more precisely as the book progresses, as will other ecological areas nested within them. I begin here in figure 1.2 by denoting the names of the communities visited earlier in the chapter, starting with the Magnificent Mile just south of the Near North Side’s Gold Coast and stretching down through the Loop and Near South Side all the way to Avalon Park. We ended our tour back up in the Near North Side but west of the Gold Coast in today’s manifestation of Zorbaugh’s 1929 “slum” in the form of “Death Corner.” The reader can examine street referents and match the names of all seventy-seven communities to figure 1.2 by navigating the following website: http://www.cityofchicago.org/city/en/depts/doit/supp_info/citywide_maps.html.

41. I define child health as the standardized combination of infant mortality rates and low-birth-weight babies—these two indicators are very highly correlated (0.81, p < 0.01). As expected based on prior research, poverty is significantly correlated (0.78, p < 0.01) with poor infant health. I examine in figure 1.2 the predicted infant health rate from its poverty rate across communities (displayed in thirds). The probability that the relationship between predicted infant health and homicide occurred by chance is less than one in one thousand and thus is highly nonrandom.

42. Another way to demonstrate the spatial link is to estimate the association of homicide with infant mortality and low birth weight while directly controlling for poverty. When I do so, the partial correlation is 0.60 (p < 0.01). Poverty thus does not explain away the clustering.

43. Population size cannot explain the patterning of events either as that is also adjusted. Interestingly, even within the highest-crime-rate community of Chicago, crime is highly concentrated by time and place, almost as if the wider citywide patterns of ecological concentration are refracted onto the microlocal level of a single block. See St. Jean 2007.

44. Drake and Cayton [1945] 1993. See especially maps on pages 203 and 205.

45. For the most thorough review of the intellectual history of the social disorganization tradition, see Kornhauser 1978. I will synthesize this literature in chapter 2.

46. These data will be described in further detail in chapter 8.

47. This is an increasingly dominant viewpoint (McCarthy, Miller, and Skidmore 2004).

48. These measures and data sources will be explained in full in later chapters.

49. Florida 2002. For an assessment of Florida’s argument in Chicago, see Graif 2010.

50. The partial correlation of rates of Internet use and bohemians across communities, controlling for median rent and median income, is 0.55 (p < 0.01). Keith Hampton also shows that residents who are digitally “wired” and frequent users of the Internet are more connected to their neighbors and more involved in local collective action than their disconnected counterparts (Hampton 2010; Hampton and Wellman 2003). Thus even the digital world is “placed.”

51. Sampson 2008a, 133. As of the mid-2000s there were only a handful of Starbucks south of Roosevelt Road in Chicago; most cluster on the North Side. A Starbucks in Wicker Park was visibly defaced with graffiti on a visit in 2007. Clashes of culture were readily apparent. Exiting Myopic Used Books, where one can buy any number of dog-eared radical treatises, a perfectly opulent woman from American Apparel stared down at me from a billboard.

52. As argued by Wacquant 2008, for example.

53. See also Sampson and Wikström 2008.

54. Wikström 1991.

55. “Lax Real Estate Decisions Hurt Starbucks,” New York Times (Stone 2008).

56. The neighborhood-effects and place stratification literatures represent important exceptions, as reviewed in chapter 2. Still, it seems that jet-set academics and newspaper columnists fall hard for the notion that neighborhoods do not matter. At least, perhaps, until they return to upscale homes in economically segregated communities with high test scores for (private?) schools. Or to the highly secure condominium with elaborate owners’ associations set up to protect and enhance neighborhood investments. It is also notable how few “the world is flat” critics buy homes in the world’s violence-ridden slums, where property comes dirt cheap. Doing so would appear to be a quite rational move if individual choice is what matters, not neighbors or neighborhoods.

57. Castells 2000, 697; 1996.

58. Sassen 2001; Massey 1996.

59. Even emotional attachment to place and local interaction, widely considered dead according to contemporary assertions, are not dead according to empirical accounts. When Albert Hunter attempted a replication twenty-five years later in the same Rochester, New York, neighborhood studied intensively in 1950, he found a decline in local facility use (e.g., shopping, entertainment) but no change in neighboring and local interaction. The “sense of community” actually increased, leading Hunter to reject the idea of a cultural-symbolic loss of community (Hunter 1975, 549). At the national level, Americans continue to derive a sense of identity from their neighborhood with over 50 percent of the population reporting that they talk with their neighbors several times a week or more (Social Capital Community Benchmark Survey 2010). It has also been argued that the widely discussed finding of a shrinkage in Americans’ personal networks stems from an error in the data (Fischer 2009).

60. Gieryn (2000, 465) presents an excellent sociological review of the concept of place. For a political science perspective, see Agnew 2002. The geographical view is iconoclastic by the standards of contemporary wisdom and more in line with this book, noting the “power of place” (De Blij 2009), not its disappearance. The intellectual history of geography is interesting in its own right. It went so far out of fashion as a discipline that its departments were shut down in several major universities. Geographical science is now in resurgence in those same universities. The social science literature on neighborhoods, a special kind of place, is voluminous and will be synthesized in the next chapter.

61. I wish to be clear that this book is not an official statement of the project as a whole. Not all participants in earlier stages would necessarily agree with my theoretical framing or analysis. Chapter 4 reveals the diversity of thought that went into the planning of PHDCN.

62. Daniel Burnham, urban planner extraordinaire, famously nudged Chicago at the beginning of the twentieth century, to “make no small plans” (Smith 2006). Nelsen Algren (1951), along with many others before and after, has noted the character of a city always on the make.

63. Small 2009; Sampson et al. 2005; Marwell 2007; McQuarrie and Marwell 2009.

CHAPTER TWO

1. One might even say an unrepentant descendant except that I believe that to adhere rigidly to any one “school” concedes too much intellectual terrain. Fortunately for the Chicago School, this risk is minimized by the nature of what it was not. Howard Becker (1999) makes the useful distinction between schools of thought (which imply unanimity of intellectual outlook) and schools of activity. If the Chicago School of urban sociology between the two world wars can be said to be a school, it was in forms of empirical activity rather than an agreed upon theoretical commitment. Enforced schools of thought are usually constructed after the fact by outsiders; see Abbott 1999.

2. Molotch 2002. Michael Dear fired the first shot in a symposium declaring an ascendant LA School 2002, with responses by Abbott 2002 and Sampson 2002a among others. For a discussion of the French School and new urban sociology, see Lefebvre 1991 and Gottdiener and Hutchison 2006. The “New York” School has been presented in Halle 2003.

3. See Laub 2004 for a discussion of the application of life-course principles to intellectual ideas. I am particularly interested in the concepts of pathways and turning points in the idea of neighborhood effects.

4. Levin and Lindesmith 1937, 801. On the repeated rediscovery of neighborhood over time, see Sampson and Morenoff 1997.

5. Morris 1958, 51. See also his review of English social ecology in general.

6. Quetelet [1969] 1842, 87.

7. Morris 1958.

8. Rawson 1839, 336.

9. Mayhew 1861.

10. Booth 1889.

11. Shaw 1930.

12. Mayhew 1862, 272.

13. Shaw and McKay [1942] 1969.

14. For searchable access to archive materials from the Booth collections of the Archives Division of the Library of the London School of Economics and Political Science and the University of London Library, see http://booth.lse.ac.uk/ (LSE 2008).

15. Johnson 2006.

16. For further intellectual history, see Abbott 1997.

17. Park and Burgess 1921. For an image of the map: http://www.csiss.org/classics/content/26. On the Green Bible, see Short 1963, xvii.

18. Park 1915.

19. Park and Burgess [1925] 1967.

20. Social workers Jane Adams and Merritt Pinckney vocalized the need to address the deteriorated housing, mental disorder, truancy, infant mortality, and delinquency they deemed a result of rapid industrialization and social change. Sophonisba Breckinridge and Edith Abbott (1912) examined the geographic distribution of cases brought before the Juvenile Court of Cook County (Chicago) on delinquency petitions and showed that a disproportionately large number of cases were concentrated in certain neighborhoods. While these early social workers were not methodologically sophisticated, their emphasis on the relationship between the volume of delinquency and local community characteristics was nonetheless influential.

21. Shaw 1929.

22. For a review of neighborhoods and health, see Sampson and Morenoff 1997 and Sampson and Morenoff 2000.

23. Goldberger, Wheeler, and Sydenstrycker 1920, 2707.

24. Sampson and Morenoff 1997.

25. Kawachi and Berkman 2003.

26. Shaw and McKay [1942] 1969, 106.

27. Faris and Dunham 1939.

28. Literally, in fact. Drake and Cayton’s maps on delinquency and insanity published originally in 1945 ([1945] 1993, 203, 205) were borrowed from the Institute of Juvenile Research where Shaw and McKay worked, and from Faris and Dunham, respectively.

29. Kornhauser 1978; Bursik 1988. Social disorganization theory is commonly traced to the work of Shaw and McKay, who in turn drew on Park and Burgess. But equally important influences were Thomas and Znaniecki (1927, introduction, chaps. 1, 3) and Wirth (1938), who argued that ethnic heterogeneity, along with size and density, undermined social integration.

30. Sampson and Groves 1989.

31. See also Janowitz 1975.

32. Whyte 1943.

33. Gans 1962.

34. Jacobs 1961; Stack 1974.

35. Putnam 2000.

36. Bursik 1999.

37. Sampson, Morenoff, and Gannon-Rowley 2002.

38. Robinson 1950. For an interesting contemporary take, see Subramanian et al. 2009.

39. Shevky and Bell 1955. In many ways the social area analysis tradition may be seen as a continuation of the typological work that Rawson promoted in nineteenth-century London.

40. See, for example, Berry and Kasarda 1977.

41. Morenoff and Sampson 1997.

42. Harvey 1973; Castells 1977; Gottdiener and Hutchison 2006.

43. Their argument culminated in the book Urban Fortunes (Logan and Molotch 1987).

44. Strip malls, restrictions on mixed-income housing, the separation of stores from residences, minimum lot sizes, gated enclaves, surveillance demands, and the banishment of sidewalks have all been blamed for undermining the authenticity of urban places. For a perceptive analysis of New York in the aftermath of Jane Jacobs, see Zukin 2010.

45. Bursik 1989.

46. Skogan 1990. Nationwide, fully 20 percent of all central-city housing units occupied by blacks were lost in the period 1960–70 because of urban redevelopment alone. This does not include displacement brought about by routine market forces (e.g., evictions, rent increases).

47. Wacquant 1993. For an argument on the “planned shrinkage” of state resources in the case of New York City, see Wallace and Wallace 1990.

48. Wilson 1987.

49. See also Massey and Sampson 2009.

50. Wilson 1996.

51. Wilson 1987, 56.

52. See also Jargowsky 1997.

53. For a critical account that claims Wilson is a conservative because he fails to properly attribute inequality to racism, see Steinberg 1995. Another critique of neighborhood-effects research for being conservative comes from a student of Wilson’s. Wacquant asserts that neighborhood effects are really “state” effects. He uses the 1996 Welfare Reform Act in the U.S. as an example of the retrenchment of the state in producing a concentration of urban “outcasts” (Wacquant 2008). Because this action was at the federal level, however, the mechanisms that explain how a national policy that was a constant across localities created neighborhood-level variations beyond those that were already in place are not explained. Macro effects are not incompatible with this book’s argument; in fact a macroneighborhood interface was part of the logic of Wilson’s original thesis and the idea of concentration effects. I directly address in later chapters the related critique that neighborhood effects research has tended to neglect cross-neighborhood linkages and the mediation of state resources by nonprofit organizations (Marwell 2007). The social isolation hypothesis that is embedded in Wilson’s work has also come under criticism for assuming a monolithic culture in the ghetto. I think this criticism is partially valid although I read into Wilson a greater emphasis on forms of spatial isolation that renders contact with diverse outlets—including mainstream values—less likely. I return to issues of culture at several points in this book.

54. Massey and Denton 1993.

55. Morenoff and Sampson 1997.

56. Massey 1990, 337; see also Massey and Eggers 1990.

57. Both also find empirical support in Quillian’s (1999) adjudication.

58. Nisbet 1953 argues, correctly I believe, that the real question is not why the decline, but rather why there is such an obsession with community decline that reinvents itself over time. For relevant insight on the fractal nature of intellectual change, see Abbott 2001.

59. Some of the most ambitious foundation efforts in the last decade include those by the Ford, Rockefeller, and MacArthur foundations to increase the capacity of local communities to solve common problems. Nongovernmental organizations (NGOs) have grown noticeably as well, many embracing similar ideas of building community. A notable example is the Community Development Corporation (CDC) movement, which has long privileged community as a meaningful unit of social intervention to improve the lives of the poor. For further elaboration of this point and especially criminal justice policy, see Sampson 1999.

60. For a review and critique of the new urbanism, see Glaeser 2011.

61. World Bank 2010.

62. Robert Putnam was a well-known advisor to Clinton and Tony Blair in the UK. For an interesting account of the Bush view of community, see Milbank 2001. There is every indication that President Obama is attuned to community issues, coming as he does from a community-organizing background on the South Side of Chicago. I return to this in chapter 8.

63. Interestingly, Nisbet’s ideology of lament seems to be pervasive among baby boomers, a group that has achieved widespread prosperity and educational achievement when considered in historical perspective. Seeking an alternative to suburban sprawl, consumerism, mainstream institutions such as old-line churches and fraternal orders, and the atomizing effects thought to stem from globalization, the baby-boom generation appears to be driving much of the demand for things like the new urbanism and a return to city neighborhoods. The NIMBY syndrome takes on new meaning here, pitting old versus new gentrifiers against one another in an attempt by the former to stop the rapid development of the community and the perceived loss of what drew their interest to restoring decaying neighborhoods in the first place.

64. This is not to say philosophical concerns are irrelevant. To the contrary, in a later section of the book I attempt to integrate normative considerations of justice (the “good community”) with empirical findings on other-regarding and socially altruistic behavior.

65. Gerald Suttles (1972b) wrote some thirty years ago on how the frenzy over “local control” and community building seemed impervious to evidence on its failures. A more recent example is the widespread popularity of community policing. For critical evidence on both community ideals and their implementation in the form of community policing, see Herbert 2006.

66. Novelists have been more astute than social scientists on the crushing weight of distorted forms of community. Dostoevsky’s Grand Inquisitor and Nietzsche are emblematic.

67. Sugrue 1996; see also Kaufman 2003.

68. Sampson, Morenoff, and Gannon-Rowley 2002.

69. The social-ecological literature has also considered neighborhood differentiation related to life-cycle status, residential stability, home ownership, density, land use, and ethnic heterogeneity. The evidence is mixed, especially for population density and ethnic heterogeneity (Morenoff, Sampson, and Raudenbush 2001; Brooks-Gunn, Duncan, and Aber 1997). There is research showing that residential instability and low rates of home ownership are durable correlates of many health-related behaviors, but there is evidence that residential stability predicts negative rather than positive outcomes (Ross, Reynolds, and Geis 2000). A more recent but understudied object of inquiry is concentrated affluence (Massey 1996; Brooks-Gunn et al. 1993). Many of these factors will be considered later.

70. Sorensen 1998, 240; Wikström and Sampson 2003. One of the earliest sources of serious sociological attention to the idea of social mechanisms was Robert Merton (1949), who focused on the self-fulfilling prophecy and the “Matthew Effect,” among other mechanisms (Hedström and Bearman 2009). The mechanistic turn in the social sciences is in its ascendancy even though there is sharp disagreement on the proper conceptual or methodological approach.

71. Sampson, Morenoff, and Gannon-Rowley 2002, 447. For the most recent assessment of the literature on neighborhood social mechanisms that I could find, see Galster 2011.

72. Jencks and Mayer 1990; Mayer and Jencks 1989.

73. Woodward 2003.

CHAPTER THREE

1. Park and Burgess 1925 [1967], 147

2. Park 1915, 579.

3. Suttles 1972, 8.

4. Choldin 1984. This in part addresses the concern of the political economy critique.

5. Suttles 1972, 59; Reiss n.d.

6. Suttles 1972, 51.

7. Hunter 1974.

8. Gieryn 2000.

9. On the concept of social homophily, see McPherson, Smith-Lovin, and Cook 2001.

10. Suttles 1972. For important work on the defended neighborhood in action against change, see Taub, Taylor, and Dunham 1987; Wilson and Taub 2006.

11. Coleman 1961.

12. Smith 2010; Smail 1999.

13. Smith 2010; Jacobs 1961, 117.

14. Warren 1975, 50.

15. Zorbaugh 1929, 65. He went on to say that in this “neighborly” sense, “there are no neighborhoods” (68).

16. For example, see Powell et al. 2010.

17. Brubaker 1996.

18. The seminal work of Suttles (1968) in The Social Order of the Slum shows how differential social organization maps onto territorial forms. See also Tilly 1973, 212.

19. Differential commitment to neighborhood might be thought of as the “community of differential liability,” a variant on Janowitz’s (1975) conception of the “community of limited liability,” where attachment to neighborhood is voluntary and often based on instrumental values surrounding investment rather than the constrained interpersonal ties that characterized the “urban villages” of our past. I accept most of Janowitz’s reasoning but resist its implication of the seemingly inevitable decline of community.

20. Wilson 1987. For a pre-Wilson analysis, see Massey and Sampson 2009.

21. See, for example, Anderson 1990, 1999; Bourgois 1995; Newman 1999; Sanchez-Jankowski 2008. Notable exceptions to the inner-city poverty restriction include Pattillo-McCoy’s exploration of the black middle class (1999,) in Chicago and Lacy’s (2007) study of the suburbs of Washington, DC. See also ethnographies by Kefalas (2003) and Carr (2006) of a white working-class area in Chicago and Katz’s (2009) comparative study of Los Angeles neighborhoods in the larger Hollywood community. Largely absent in this tradition are studies of mixed or Latino (but see Small 2004; Klinenberg 2002; Wilson and Taub 2006; Katz 2009) and white upper-income areas.

22. Sanchez-Jankowski 2008, 9. See also Pattillo 1998.

23. Somewhat surprisingly, upon close reading one finds that it is not unusual for urban ethnographies to make comparative or even causal claims about neighborhood effects even though empirical observations are within one place. To be clear, however, I am not critiquing the empirical or scientific validity of ethnography, but rather specifying the logic of specific claims.

24. See also the methodological arguments and ethnographic work by Katz 2001, 2009.

25. Firey 1947 is usually considered one of the first to systematically tackle cultural issues in the study of urban ecology.

26. At the level of the nation-state, the classic statement is Benedict Anderson [1983] 1991, Imagined Communities. For a fascinating historical study of neighborhoods and the mental use of maps in medieval Marseille, see Smail 1999.

27. Gieryn 2000, 465. See also Katz 2009.

28. Coleman 1994. See also Clampet-Lundquist and Massey 2008.

29. Thus psychometrics turns on the measurement of individual differences, ecometrics on ecological or neighborhood differences. See Raudenbush and Sampson 1999.

30. Park and Burgess [1925] 1967.

31. Wilson 1987.

32. Janowitz 1975; Marwell 2005, 2007.

33. See also Coleman 1986. Again, my consideration here does not signal a commitment to methodological individualism or rational choice theory, nor, I think, was Coleman intent on claiming that there was only one way to study micro-macro transitions.

34. Economists are often viewed as being at the forefront of the selection-bias concern, but the sentiment is widely shared in the social sciences, with the seminal critique of neighborhood-effects research offered by Jencks and Mayer (1990).

35. Heckman 2005.

36. The empirical forays of methodological individualists betray a reductive emphasis on individuals and lack of deep concern with social context in its own right (Abbott 2007; Jackson and Pettit 1992). A possible exception is “analytic sociology” and the associated concept of “structural individualism” (see Hedström and Bearman 2009), although the “supervenience” assumption and associated epistemological stance of analytic sociology has, in my opinion, deflected attention from the ontology of social-level or contextual causality (see chapter 15). Although for different reasons, the classic perspective on ecological context in developmental psychology, that of Bronfrenbrenner (1979), has also in practice focused on the individual despite the theory’s emphasis on the “mesosystem” and macrosystem.

37. Sawyer 2005. See also Hedström and Bearman 2009.

38. This framework extends my initial argument in Sampson 2002a combined with my reading of a long line of urban scholarship mainly from World War I to the present. It also builds in part on the vision laid out by the late James Coleman before his death (Coleman 1994, 1986) and Andrew Abbott’s intellectual history of Department and Discipline (Abbott 1999).

39. Snow 1993.

40. Chapter 2; Abbott 1997, 1152. Such a stance does not rule out generalities, only a fine balancing act between treatment of the city as both lab and social field (see chapter 4). Although not a terribly pleasing phrase, the concept of the “middle range” (Merton 1949) has much to offer in the way of eliding the unhelpful divisions that still rile much of social science, such as basic/applied, theory/research, and positivist/interpretive (Sampson 2010a).

41. We have made virtually all of the major components of the PHDCN data public over the years, within the constraints of confidentiality agreements (see http://www.icpsr.umich.edu/icpsrweb/PHDCN/). Over a hundred scholars have already published papers on the data that we are aware of. There are likely more papers not reported to ICPSR, and others will appear in the future. I will also be making much of the data from this book available through ChicagoMap, a project of the Center for Geographic Analysis at Harvard (see chapter 15).

42. See Burawoy 2003 for his notion of the “revisit” to ethnographic field sites. On an abstract level, one might say here I attempt a targeted revisit to multiple Chicago communities, a comparative reassessment using an assortment of methods, but mostly quantitative, of neighborhoods that have been studied over the years. My views have evolved as a result of this process, most notably an increased appreciation for historical and cultural processes.

43. Appendices are a common device to present such material, but they tend to reify a false distinction between the main line of theoretical argument and analytic methods. I thus eschew the strategy of multiple appendices and integrate methodological and theoretical arguments into the ongoing flow of the presentation, with additional details provided in footnotes.

CHAPTER FOUR

1. The interpretation of PHDCN offered in this chapter is my own and no doubt is shaped by my intellectual history. Although I make every effort to be accurate and I conducted multiple interviews in piecing together the story, others may of course have seen things differently at the time. I also apologize in advance if not all people with roles in the project are sufficiently recognized in print. This was truly a collaborative project with literally hundreds of participants.

2. The Framingham Heart Study (FHS), conducted in the place known as the “The Town That Changed America’s Heart,” has long been considered the gold standard in research on health. The FHS was a clustered community-based study that started some fifty years ago and that has followed the original participants over time with intensive measurement. Its influence has been enormous (Framingham Heart Study 1995–2006). The PHDCN, as we shall see, purposefully eschewed the “national sample” approach of most surveys and comes closer to the Framingham ideal of in-depth scientific assessment in a place thought to be broadly representative of the types of problems facing Americans.

3. Cloward and Ohlin 1960; Wilson and Herrnstein 1985. Norval Morris was also important, but my take is that he played more of an administrative rather than intellectual role in the project’s design. Although not on the original committee, David Farrington of the University of Cambridge was a major intellectual influence behind the idea of a multicohort longitudinal design, as discussed below.

4. This tension has roiled the study of crime since the days of Cesare Lom-broso and the search for the “born criminal” or “criminal man” in the 1800s. He is still being sought. For a new translation of this influential Italian criminologist, see Gibson and Rafter 2006.

5. Wolfgang, Figlio, and Sellin 1972.

6. Published as Laub 2004.

7. The intense intellectual focus in criminology on prediction follows an interesting path over the course of the last century, a topic John Laub and I explored in “The Sutherland-Glueck Debate: On the Sociology of Criminological Knowledge” (Laub and Sampson 1991). See also Laub 2004.

8. March 31 (p. 1) and November 28 (p. 1), 1994, respectively.

9. Farrington, Ohlin, and Wilson 1986.

10. For a fascinating intellectual history of the Chicago School, including its colorful characters, see Abbott 1999.

11. Reiss and Tonry 1986.

12. The late David Rowe and I spent many hours disagreeing but always with pleasure.

13. For an early case for why a new longitudinal study was needed, see Farrington, Ohlin, and Wilson 1986. For a detailed description of design-related features that went into the planning phases of the PHDCN, see Tonry, Ohlin, and Farrington 1991.

14. The composition of the leadership team changed over time. For example, the psychologist Terrie Moffitt left fairly early on to take up work on another longitudinal study, and Jeanne Brooks-Gunn joined near the end to represent developmental psychology. Elizabeth Susman and Jeff Fagan also helped out in the middle phase, and the late Bob Cairns was a key figure in the early stages of the project’s developmental thinking, as was Michael Tonry in the management of the group’s early planning activities. I was scientific director for the community design, and Stephen Raudenbush was scientific director for analysis.

15. See also Earls and Buka 1997.

16. Judt 2007. Mailer called New York “one of the capitals of the world.”

17. As David Brooks (2009) notes: “All smart analyses of the Obama administration begin with Chicago.”

18. Kotlowitz 2004.

19. Whereas NYC, LA, and Chicago, to name the big three, certainly look different, we need to distinguish between surface difference and underlying fundamentals on neighborhood inequality and social processes. Consider Latino immigration. Readers might be surprised to learn that metropolitan Chicago has almost 1.5 million residents of Latino or Hispanic origin, by some estimates the second-largest concentration in the country. The city proper grew during the 1990s and is one of the most ethnically diverse cities in the United States. One might be further surprised that Hispanic-white segregation (the index of dissimilarity, ranging from 0 to 100) is 63 in LA compared to an almost identical 62 for NYC and 67 for Chicago. Asian-white segregation levels are also similar, 48 for LA compared to 50 for NYC and 44 for Chicago. Or consider that violence follows a general pattern of ecological concentration in LA just as in other cities. Solid empirical research on changes in crime in LA County over recent decades has demonstrated a timeworn social-ecological pattern linking neighborhood deterioration and rising crime rates in late twentieth-century LA to concentrated disadvantage and associated neighborhood factors that have predicted crime distributions in American cities going back to early twentieth-century Chicago. It is true that some forms of race-linked segregation are new to American cities in the post-Fordist era (and remain more entrenched in Chicago), but a large research literature has attacked this very issue. Note as well that Mike Davis’s (1992) fiery depiction of the coming ruin of Los Angeles runs opposite to the facts. Crime is down and the city is booming like many others. In short, while LA and NY look and feel different than Chicago, and while the LA School is clearly right about the distinctive urban form and ecological routines in LA, the question is how much this matters for the sorts of questions I pose.

20. One of the best papers highlighting the struggle to balance field and laboratory approaches, with Chicago as the case study, is Thomas Gieryn’s “City as Truth Spot” (2006). In the present case, I go back and forth in using Chicago as both a field and laboratory.

21. Ever in aggressive overdrive, the headlines in the Chicago Tribune when I first drafted this chapter extolled the can-do virtues and American naturalness (versus Los Angeles “phony”) that would win Chicago the 2016 Olympics. A few weeks later Chicago was selected as the American representative over LA. Chicago grit again won out, but in the end Rio was the charm.

22. For a history of Chicago community areas, see Chicago Fact Book Consortium 1990. See also Hunter’s (1974) important analysis of neighborhood and community identity in Chicago.

23. The Latino community of South Lawndale is known locally as “Little Village,” for example, but the area remains distinctive and set apart from North Lawndale. Similarly, the Lower West Side is widely known as “Pilsen.”

24. Suttles 1972. See also Hipp’s (2007) analysis of variability in ecological definitions and Grannis’s (1998, 2009) description of “tertiary communities” that takes into account the layout of streets to comprise areas that are connectable without crossing major thoroughfares.

25. Black poverty is well known. That there were no white high-poverty neighborhoods is less well appreciated. High-income Latino and high-income Latino/African American neighborhoods were also nonexistent (Sampson, Raudenbush, and Earls 1997, 919). In this sense the research design itself produces a striking comment on segregation by race and class in Chicago. But the same finding has been replicated in other cities, suggesting that Chicago is not unique.

26. There were over fifty research assistants who carried out screenings and interviews, many in Spanish and Polish. The staff represented a diversity of languages and many were immigrants themselves, reflecting the wide diversity of the Chicago area (Holton n.d.). There were supervisors and separate units for data collection, official records, community relations, tracking, and statistical analysis. The staff was recruited for the long run, with many research assistants sticking with the same families over the full period of the study, enhancing the participation rate. John Holton was the original site director of the PHDCN; he left in 1999. Alisú Schoua-Glusberg was the director of survey operations and oversaw a revised organizational structure. The manager of field operations was Kelly Martin, the data research manager was Nancy Sampson, and the administrative manager was Cynthia Coleman.

27. For further details on the PHDCN data collection for the cohort study, see Martin 2002.

28. Among the scholars in attendance besides PHDCN investigators were Douglas Massey, Claude Fischer, Richard Taub, Ralph Taylor, Lloyd Ohlin, Mercer Sullivan, and Terry Williams.

29. The Community Survey had three stages. At stage 1, city blocks were sampled within each NC; at stage 2, dwelling units were sampled within blocks; at stage 3, one adult resident (eighteen or older) was sampled within each selected dwelling unit. Abt Associates carried out the screening and data collection in cooperation with research staff of PHDCN, achieving a final response rate of 75 percent. The design produced a representative probability sample of Chicago residents and a large enough within-cluster sample to create reliable between-neighborhood measures. The samples within NCs were designed to be approximately self-weighting, and thus between-neighborhood analyses in this book are, unless otherwise noted, based on unweighted data (Sampson, Raudenbush, and Earls 1997, 924). Descriptive data are weighted. The design of the CS allows me to construct measures at multiple geographic scales, including block group, census tract, NC, and community area. In supplemental work I defined “tertiary communities” (see Grannis 2009) made up of aggregations of block groups wherein streets were connected without crossing major thoroughfares. This definition resulted in 221 areas slightly larger than neighborhood clusters defined in the PHDCN (about eleven thousand residents on average) and that produced virtually equivalent ecometric properties for the community survey measures. In the analyses of this book I rely mainly on three sets of units— census tracts (N = 875), NCs (343), and community areas (77). This provides the largest contrast in sample size and boundary definition, with the middle unit (NCs) most conforming to the PHDCN design.

30. The term “ramping up” was used often by the initial supervisor of data collection, but, as it turned out, more in terms of aspirations than a description of actual progress. He was soon fired by the contractor.

31. For a detailed sociological analysis of the Chicago heat wave, see Klinenberg 2002.

32. I am indebted to John Holton, who was PHDCN site director from 1993 to 1999, for his field notes, permission to quote, comments on this chapter, and recollection of these and other events as described in a personal interview (conducted March 22, 2007).

33. At about the same time I witnessed an ugly event at a University of Chicago seminar chaired by Christopher Jencks. James Q. Wilson, an early advisor to PHDCN, was slated to give a formal presentation. Before he could start, about a half dozen protestors stood up and starting shouting and chanting that Wilson was a racist in favor of genocide. His work with Richard Herrnstein in Crime and Human Nature and on the PHDCN was cited by the protestors. The university police were called and the protesters eventually removed, but as I recall not before the seminar had to be aborted. The politics of race and crime have always been deeply controversial, but the early to mid-1990s was an especially volatile time. In this case, to the best of my recollection, Wilson that day was not even scheduled to talk about race and crime. See Massey and Sampson 2009 on the scholarly politics of race that became a legacy of the Moynihan report.

34. The pragmatics and cost of collecting biological measures, using the technology available in 1995, were the driving reasons the PHDCN did not move forward in this area, although the protests were not insignificant either. In addition we concluded that there was a lack of scientific consensus that biological differences, such as they could be measured reliably at the time, were major predictors of crime. It is surprising how fast some things change—today the drawing of blood and obtaining genetic markers is routine.

35. During a discussion of behavioral genetics with a biologist at a professional meeting at the time, he half jokingly asserted that what is needed is a “Department of Molecular Sociology.”

36. With the growing use of cell phones and difficulty of interviewing households in person, combined with the increasing constraints imposed by our human subjects’ review boards, I nonetheless worry about the future of big social science.

37. Abbott 1997.

38. Reiss 1971, 4.

39. NORC 1995. As a check on quality control, new observers recoded a random 10 percent of all coded face-blocks, and the results compared. This test produced over 98 percent agreement.

40. Short 1963, xvii, in his introduction to Thrasher 1927 [1963].

41. The probability sample was nested within the NCs defined by PHDCN originally, with oversampling within the eighty focal NCs (of cohort selection). The 3,105 interviews completed include 1,145 respondents in focal NCs (37 percent of the total sample or just over 14 respondents per NC on average), and 1,960 in nonfocal NCs (63 percent of the total sample, or just over 8 respondents per NC on average). For purposes of studying racial-ethnic disparities, the sample is very balanced, including 802 Hispanics, 1,240 non-Hispanic blacks, 983 non-Hispanic whites, and 80 people of other races/ethnicities. The second community survey yielded a response rate of 72 percent, close to matching the first CS even though personal surveys became harder to conduct over time.

42. To ensure that interviewers made reliable and valid observations, ISR conducted a pilot study in which two raters made independent observations of the same block at the same time for eighty blocks, one in each of the focal NCs. In the second stage of the SSO, ISR collected observations from a single rater on almost every block in Chicago that contained a CCAHS survey sample address. All together 1,664 of the total of 1,672 blocks were assessed (approximately eight blocks per focal NC and four blocks per nonfocal NC), for a total of 13,251 face-blocks nested within 1,379 census block groups, 703 census tracts, and 343 NCs.

43. We also sampled new organizations and determined where occupants of the original leadership position went. The 1995 Key Informant study was more directly linked to the PHDCN community design. The second panel study in 2002 was a cognate effort independently funded by the American Bar Foundation, the MacArthur Foundation, and the Chicago Community Trust.

44. I was also a Research Fellow for seven of these years at the American Bar Foundation, located on the Near North Side.

45. See also Gieryn 2006, 24.

46. Becker 1996.

CHAPTER FIVE

1. Drake and Cayton [1945] 1993; see the maps on pages 203 and 205.

2. Moynihan 1965. Kenneth Clark was a contemporary of Moynihan who made his own provocative splash in Dark Ghetto (1965), apparently originating the “pathology” metaphor.

3. On the intellectual and historical context, see Massey and Sampson 2009.

4. Properly so, I would argue, for pathology too easily conjures up the medical model as the correct paradigm to follow. Criticisms of terms like “pathology” and “disorganization” are noted further in chapter 7. Note, however, that the logic used by as Moynihan was not restricted to the black community. The manifestations of pathology were instead thought to vary by location in the social structure. Thus while the terminology is certainly problematic for social phenomena, we should not lose sight of the structural logic that framed Moynihan’s argument.

5. This chapter extends the thesis in Sampson 2009b.

6. Moynihan 1965, chap. 3, p. 1. On variable-based approaches, see Abbott 1997.

7. Bowles, Durlauf, and Hoff 2006; Sampson and Morenoff 2006.

8. Moynihan 1965, chap. 5, p. 1

9. Shaw and McKay 1942; Drake and Cayton [1945] 1993; Wilson 1987; Massey 1990.

10. Sampson, Morenoff, and Gannon-Rowley 2002; Land, McCall, and Cohen 1990.

11. See chapter 4 on the pros and cons of census tracts as neighborhood units of analysis.

12. Sampson, Sharkey, and Raudenbush 2008, 848, table 1.

13. Based on this result, I drop density of children from the scale in later analyses. When I examine concentrated disadvantage without racial composition, the resulting scale correlates 0.99 with the initial scale that included percentage black. Results are thus robust across definitions.

14. The full distribution is shown in Sampson 2009b, 266.

15. For further elaboration, see Sampson, Sharkey, and Raudenbush 2008.

16. Massey and Denton 1993.

17. The correlation is 0.59 (p < 0.01) in mixed areas. In black areas it is even larger, at 0.78 (p < 0.01). For a chart of this relationship, see Sampson 2009b, 267.

18. Bureau of Justice Statistics 2010.

19. Western 2006.

20. Sampson and Loeffler 2010.

21. We measure the imprisonment rate by the number of unique Chicago-residing felony defendants sentenced to the Illinois Department of Corrections from the Circuit Court of Cook County divided by the number of Chicago inhabitants ages eighteen to sixty-four in the 2000 census. We use adults “at risk” in the denominator in order to rule out confounding variations in the prevalence of children under age eighteen that are not eligible for prison. Similarly, we exclude those at the older end because, while society is aging, very few prisoners are over sixty-five. These data were gathered from the electronic records of the Circuit Court of Cook County (Sampson and Loeffler 2010).

22. Wilson 1987. See also Jargowsky 1997.

23. Koval et al. 2006.

24. Sampson and Morenoff 2006.

25. Wilson and Taub 2006; Hunter 1974.

26. Suttles 1990.

27. See also Taub, Taylor, and Dunham 1987.

28. Massey, Condran, and Denton 1987.

29. Dust jacket of Hyra 2008.

30. The sociologist Malcom Klein, in remarks at a plenary session on the Project on Human Development in Chicago Neighborhoods, American Society of Criminology, 2006. The irony is that Klein is from Los Angeles, not the first city that comes to mind as unweird.

31. Peterson and Krivo 2010.

32. Small 2007; Small and Feldman 2008. Some have also claimed that black neighborhoods in Chicago are “remarkably different than those in the average American city” (http://www.columbia.edu/^~jwp70/COMURB_Ghetto-Chicago_discussion_2008.pdf). See also the follow-up symposium in the journal City and Community 7 (2008): 347–98.

33. Sampson 2009b, 268. I used a principal-components, regression-weighted scale of the indicators introduced above: poverty, female-headed families, welfare assistance, unemployment rate, and racial isolation (percent black). I also used census tracts as the definition of neighborhood (chapter 4) to provide a U.S. comparison.

34. See also Sampson 2009b.

35. Peterson and Krivo 2010.

36. For a scholarly assessment, see Blumstein and Wallman 2000.

37. See http://www.census.gov/acs/www/Products/PUMS/. I used data from the period 2005–7 to increase precision of estimates, as it contains records for about 3 percent of housing units in each geographic area compared to about 1 percent for the one-year files.

38. http://www.uic.edu/cuppa/voorheesctr/Publications/vnc_woodsrpt_0706.pdf, page 9.

39. Wilson 1996.

40. For a chart, see Sampson 2009b, 271.

41. I estimated the equation: Poverty 2000 = a + β*Poverty 1960 + ε. The expression (α + β*Poverty1960) is the 2000 predicted poverty rate and ε is the deviation or residual from the predicted rate. The residual reflects the amount of change in a given neighborhood’s poverty that is unaccounted for or unexplained by its initial level of poverty. Residual change scores have the advantage over raw change scores of being statistically independent of initial levels of poverty. In addition, because all of Chicago was used to estimate the regression equation upon which the residuals are computed, these change scores have the desirable property of incorporating the dynamics of the entire ecological structure (see also Bursik and Webb 1982). The 0.91 correlation between 2000 and 2005–9 poverty at the community-area level means that the pattern in figure 5.6 is virtually identical when the latter is substituted. However, because the 2005–9 data were not available until late 2010 and the sampling is different from previous census decades, I retain the 2000 poverty rate to measure residual change.

42. Not examined here is the growth in suburban poverty in southern Cook County. Towns such as Harvey, Dolton, Country Club Hills, and other south suburbs saw growth in poverty and after the CHA transformation, in-migration from former residents of public housing.

43. Koval et al. 2006.

44. Pattillo 2007; Hyra 2008. I will return to these areas in the final chapter.

45. A recent study of the U.S. reports a robust relationship between income inequality and income segregation, but the effect is larger for black families than it is for white families (Reardon and Bischoff 2011).

46. Moynihan 1965; Clark 1965.

CHAPTER SIX

1. Jacobs 1961.

2. Booth 1889; Charles Booth Online Archive 2010.

3. Quoted in Pfautz 1967, 191.

4. Economist 2006, 57–58. My discussion here draws on Sampson 2009a.

5. Sennett 1970. See also his reflections some thirty years later (Sennett 2009).

6. Benedict Anderson later wrote on the “imagined community” and identity formation at a “nation” level, but in broad theoretical terms the mechanisms are similar [1983] 1991.

7. Hunter 1985.

8. Goffman 1963a, 9.

9. Lofland 1973, 22, emphasis in original.

10. Putnam 2007.

11. Skogan 1990, 75.

12. Kelling and Coles 1996, 108–56. For a novel counterpoint, see Duneier 1999.

13. Economist 2006.

14. Jordan 2005.

15. Ross, Reynolds, and Geis 2000; Geis and Ross 1998.

16. Sampson and Raudenbush 1999.

17. This framework does not imply that disorder is unimportant for explaining neighborhood social dynamics. Although crime and disorder may reflect common origins, crime may be less relevant for understanding urban processes such as population abandonment because it is largely unobserved. Corresponding to the “text” from which all actors in a neighborhood read, disorder is a visually proximate and immediate neighborhood cue of theoretical interest, even if it is a not a direct cause of further crime. I return to this thesis in a later section.

18. This is an important discovery, as individual reliability (e.g., perceptions of disorder) is not the same thing as the reliability of a measure to detect between-neighborhood variance. For further discussion of this methodological issue, see Raudenbush and Sampson 1999.

19. The reason for the robbery link, we proposed, is that disorderly areas serve up vulnerable potential victims, such as drug dealers, drunks, and customers of prostitutes. For a fascinating discussion and analysis of the ecology of disorder and robbery, see St. Jean 2007.

20. I set aside for now further consideration of the somewhat narrow question of whether broken windows “cause” crime. Current data cannot resolve the question satisfactorily, and in any case, detailed reviews of the literature are available elsewhere (e.g., Harcourt 2001; Skogan 1990; Taylor 2001; St. Jean 2007). However, I would note here an interesting study that appeared in Science in 2008. Researchers using a set of field experiments examined the “spreading of disorder” by manipulating aspects of physical disorder (e.g., graffiti, litter) to see if passersby in the city of Groningen, Netherlands, were influenced to act inappropriately. Keizer, Lindenberg, and Steg (2008) report that “micro” settings of disorder significantly increased the adoption of analogous disorderly behaviors and what they deem stealing. For example, Netherlanders were more likely to litter when collecting their bicycles if the bike park area was covered in graffiti than if it was clean. They were also more likely to open or take an envelope that visibly contained a five Euro bill meant for the mail if the postbox on which it was placed by experimenters was in a disordered condition. Disorder begets disorder, they concluded. The broken windows hypothesis put forth by Keizer, Lindenberg, and Steg is that disorder is a descriptive norm that undermines the tendency of humans to follow injunctive norms of appropriateness. Thus if I see a lot of litter it works to inhibit my resolve and may spread to other deviance. While clever, this study does not explain why disorder is concentrated rather than spread over geographic areas, nor does it demonstrate that disorder spreads to predatory crime, the intended target of the broken windows theory. The study also does not extend to what we might think of as “permanent” disorder. The manipulated disorder of the experiment was temporary, and if it occurred in otherwise stable areas, which seems likely, then subjects may have reasonably assumed their litter would be cleaned up. In this sense it may well be that people litter in temporary situations where there is graffiti, or that people drink more in settings where others are acting wild. Thus the assessment of the neighborhood must be considered as well, and committing murder, burglary, or robbery as a result of littering or signs of decay is another link altogether, one that remains unproven.

21. Ross and Mirowsky 1999, 414.

22. Goffman 1974.

23. Bottoms and Wiles 1992, 16.

24. Lamont 2000.

25. Loury 2002, 17.

26. For a discussion of Bayesian thinking, see Rosenkrantz 1977.

27. Ariely 2008; Thaler and Sunstein 2008. On how the pursuit of goals operates outside conscious awareness, see Custers and Aarts 2010.

28. Fiske 1998.

29. Bobo 2001; Quillian and Pager 2001; Rumbaut and Ewing 2007.

30. Wilson 1987; Massey and Denton 1993; Skogan 1990.

31. Loury 2002.

32. Fiske 1998; Banaji 2002.

33. Devine 1989; Bobo 2001, 292.

34. Correll et al. 2002, 1325.

35. Werthman and Piliavin 1967.

36. Irwin 1985.

37. Goffman 1963b.

38. Hagan 1994, 150.

39. Stinchcombe 1963.

40. Wacquant 1993.

41. Kefalas 2003, 11, 14, 43, 62, 74.

42. I am no exception. One morning a few years ago I stepped outside my home only to notice a fresh swath of painted graffiti on the wall of a nearby apartment building. My first reaction was anger and, I admit, an almost instantaneous fear that my wife and I had bought in the “wrong” neighborhood, one about to decline. But realizing that I lived in a stable, well-off neighborhood despite its dense urban character and proximity to a park and public transit line, I relaxed. I talked with authorities about the defacement as did others and the graffiti was cleaned up. Soon after, the same thing happened again and the process was repeated. After a cycle of four to five episodes, the problem went away. In this instance, “broken windows” were temporary and led to collective action, not crime or decline (Sampson and Raudenbush, 1999, 638), but the experience nonetheless taught me a lesson in the subjective emotions that disorder can inspire.

43. Laub and Sampson 1995.

44. Watt 2006.

45. Sampson and Raudenbush 2004.

46. We assessed in detail the robustness of all results (Sampson and Raudenbush 2004, 333–34, 339–40). For example, we measured and accounted for time of day in all analyses. Not surprisingly, physical disorder was highly stable over time. Even if some social disorder emerged at night (e.g., a bar fight), our results would be overturned only if it occurred in a large number of areas where other social disorder was not present during the day. From all we know on the basis of prior research and our knowledge of Chicago, such a reversal of pattern is highly unlikely. Spatial mismatch is another concern. Suppose that a resident, when responding to questions about disorder, recalled an area different from the block group where he or she lived. Our measures, however, reflected the block group as a whole and not differences within or outside block groups in the degree of observable disorder. The research design also produced a representative survey sample of individuals within block groups. Averaged across multiple residents, idiosyncratic definitions would not produce a systematic influence of racial composition on block-group variations. Furthermore, results were insensitive to controls for spatially correlated errors and variations in the size of neighborhood units, yielding similar patterns up to the community areas of Chicago, which average almost forty thousand residents. The strong similarity of findings implies that spatial mismatch cannot account for the effects of race/ethnic composition. The large magnitude of race/ethnic effects, especially in models in which physical disorder is measured virtually without error, undermines any claim that our results are artifacts of the unreliable measurement of observed disorder.

47. I address these objectives by combining the data from Sampson and Raudenbush with a panel study of the same neighborhoods. The first component of the panel is the multistage probability-based community survey (CS) based on interviews carried out in 2001–2 by the Institute of Social Research based at the University of Michigan with a new sample of 3,105 Chicago residents living in the same neighborhoods as the 1995 study. The core interview schedule from 1995 was repeated and augmented with additional questions. The design is thus a repeated cross-sectional survey that is well equipped to measure stability and change at the neighborhood level and that is representative of Chicago. The second study is also a repeated cross-section, but this time based on systematic social observation (SSO) of all block faces within the neighborhoods within which the 1995 and 2002 community survey residents lived. Based on successful results from a pretest and to save on costs, observer logs, rather than videotapes, were used on over 1,500 block groups (about seven hundred census tracts). The main SSO disorder measure in 2002 is based on the following items observed on the streets: cigarette or cigar butts; garbage/broken glass; empty bottles; painted-over graffiti; gang graffiti; other graffiti; tagging graffiti; political message graffiti; abandoned cars; condoms; and drug paraphernalia. The minimum level of reliability obtained was 0.93 at the block-group level, with 1 the theoretical maximum. Observed physical disorder is thus highly reliable at all neighborhood levels. The third data source is an integration of the 2000 U.S. census with Chicago police records on crime from 2000 up to 2006.

48. Results are similar when substituting percent foreign born for Latino. This is not surprising, as these two indicators are correlated over 0.70 (p < 0.01) at the block-group level.

49. See chapter 7 for further variable definitions. The model estimated was: Person-Level: 2002 Perceived Disorder = B0 + B1*(Age) + B2*(Male) + B3*(Black) + B4*(Latino) + B5*(Other Race) + B6*(1st Gen. Immigrant) + B7*(2nd Gen. Immigrant) + B8*(Single) + B9*(Married) + B10*(Separated/Divorced) + B11*(Kids) + B12*(Education) + B13*(HH Income) + B14*(Own Home) + B15*(Years in Neigh.) + B16*(# Moves) + B17* (# Languages) + B18*(Perceived Collective Efficacy) + B19*(Neighborly Exchange) + B20*(Perceived Violence) + B21*(Friend/Kinship Ties + B22*(Organizational Participation) + B23*(Trust in Neighbors) + B24*(Personal Victimization) + B25*(Fear) + R.

Neighborhood-Level: 2002 Adjusted Perceived Disorder = G00 + G01*(2000 Density) + G02*(2002 SSO Disorder) + G03*(2002 SSO Physical Decay) + G04*(2000 Poverty) + G05*(2002 % Black) + G06*(% Latinos) + G07*(1995 Socially Perceived Disorder) + Uo.

Equations were estimated simultaneously using hierarchical models with robust standard errors that include random individual and neighborhood components (for details, see Sampson and Raudenbush 2004). Person-level covariates are grand mean centered to capture compositional differences when assessing neighborhood-level coefficients. Twenty-four percent of the variance is explained at the individual level and 88 percent across block groups. The t-ratios (coefficient/standard error) for percent black and percent Latino are 6.67 and 9.18, respectively. The t-ratio for the black-versus-white contrast within neighborhoods is –3.66; for first generation versus third, –6.88 (all p < 0.01).

50. For example, for survey respondents nested within census tracts, blacks and Latinos were significantly less likely than whites (p < 0.01) to report disorder. First-generation immigrants scored almost a half standard deviation lower (0.19, p < 0.01) on perceived disorder than the third-plus generation. It is only at the much larger community-area level that we see racial differences within communities diminish after racial composition is controlled. This finding is not surprising because segregation between blacks and whites within community areas can be considerable. Immigrant differences within community areas also diminish but remain significant; all else equal, first-generation immigrants perceived about one-third of a standard deviation less disorder within the same community area than third-generation immigrants.

51. The original study showed an ethnic interaction in that Latinos were significantly more likely to perceive disorder as percent black increased compared to blacks and whites (Sampson and Raudenbush 2004, 335). We did not have a direct measure of generational status in 1995. The magnitude of interaction diminished over the period of follow-up study, which makes sense if we consider the argument of Loury (2002) that it takes time for immigrant attitudes to adjust. And in fact, there is no interaction of immigrant status with either percent Latino or foreign born.

52. Because of the repeated cross-sectional survey, the respondents in 1995 are different than in 2002, a crucial design element that ensures independence of neighborhood measures. The between-area or aggregate reliability of perceived disorder is 0.91 in the 2002 survey for community areas and 0.64 for tracts. See Sampson and Raudenbush 2004 for 1995 statistics.

53. For the block-group and tract level, coefficients for collective disorder were significant at the 0.01 level (t-ratios of 7.43 and 4.39, respectively).

54. All three variables were standardized to have a mean of 0 and a standard deviation of 1 before the analysis and therefore are directly comparable: the estimated effect of collective priors is more than double the effect of racial composition (0.15 vs. 0.07).

55. N = 47 communities. Over 40 percent and 97 percent of variance explained at the individual and community levels, respectively, in both the full and reduced sample model based on 1995 SSO.

56. Putnam 2000. Surprisingly, however, crime is not a major factor in his account.

57. See Merton’s (1968) original discussion of the mechanism of self-fulfilling prophecy.

58. Bursik 1986, 73.

59. Skogan 1990; Morenoff and Sampson 1997; Sampson 1986; Liska and Bel-lair 1995.

60. To account for the nonlinearity, I transformed percent poverty in 2000 into logit form.

61. One of these, Washington Park, was targeted for improvement if Chicago landed the 2016 Olympics. The city failed in its bid, but the choice gives an indication of the widely shared perceptions of Washington Park’s desperate economic and social conditions.

62. More specifically, 68 percent of the variance in the logit of 2000 census tract poverty is explained by just six neighborhood characteristics measured at a prior time. The standardized coefficients (raw coefficient divided by standard deviation) for the four significant predictors are, in order of magnitude, 0.35 (percent black in 1990), 0.33 (collective priors on disorder in 1995), 0.32 (1990 percent poverty), and 0.30 for 1990 percent foreign born. Interestingly, the two predictors that failed both substantively and in terms of significance were prior observed disorder and the homicide rate. The results thus confirm that collectively perceived disorder is on par with prior poverty and key demographics in predicting the poverty outcomes at the neighborhood level.

63. With correlations of 0.60 and 0.33, respectively. I also controlled for residential stability and racial composition within the race/ethnic strata, along with prior poverty. Collectively perceived disorder had a significant and larger independent association with later poverty in black and Latino areas (0.35 and 0.26, respectively; p < 0.01) than in white areas (0.10, not significant).

64. At the tract level for relative change, for example, the standardized coefficient for disorder as perceived in 1995 was –0.49 (t-ratio = –4.70, p < 0.01) compared to observed disorderly conditions (B = 0.09, t-ratio = 1.00). At the community-area level there is much more correlation among predictors and thus less precision in estimates along with reduced statistical power. The main pattern of results was similar, however, with shared perceptions of disorder a significant predictor of an area’s losing population. The perceived disorder coefficient was –0.35 (p < 0.05).

65. A recent examination of household moves and racial transition supports the general hypothesis of this chapter. Using a national study of households, Hipp (2010) reports that whites who perceive more crime in the neighborhood or that live in blocks with more commonly perceived crime are more likely to move out of such neighborhoods. Moreover, whites are significantly less likely to move into a housing unit in areas high in perceived crime, but African American and Latino households are more likely to move into such areas.

CHAPTER SEVEN

1. Bursik and Grasmick 1993, 30.

2. Reiss 1986.

3. Bursik 1988.

4. Warner and Rountree 1997; Sampson and Groves 1989.

5. Wilson 1996.

6. Stack 1974.

7. Pattillo 1998. See also Pattillo-McCoy 1999.

8. Venkatesh 1997. See also St. Jean 2007.

9. Browning, Feinberg, and Dietz 2004.

10. William F. Whyte’s Street Corner Society (1943) famously asserted that the problem of Cornerville was not that it was disorganized, but that its form of organization clashed with the dominant norms of middle-class society. For a recent essay on differential social organization, communities, and crime that I see as broadly compatible with my approach, see Matsueda 2006.

11. Granovetter 1973.

12. Bellair 1997. For evidence on health, see Kawachi and Berkman 2003.

13. Sampson, Raudenbush, and Earls 1997. See Bandura 1997 on self-efficacy theory.

14. In fairness it is important to note that social disorganization theorists did not equate disorganization with “chaos” or lack of patterning, rather they meant that social controls were variably aligned with the goals, needs, and resources of residents. Hence, misalignment meant disorganization. For a trenchant defense of social disorganization theory, see Kornhauser 1978.

15. Portes and Sensenbrenner 1993, 1323.

16. Hardin 2002.

17. The political scientist David Laitin suggested to me the connection of collective efficacy to the notion of “common knowledge.” Usually traced to the philosopher David Hume, the argument of common knowledge is that a fundamental condition for coordinated activity is that humans know what behavior to expect from one another. Without this requisite mutual knowledge, or what I would call shared expectations, mutually beneficial conventions would disappear. Understanding the conditions that lead to common knowledge is critical. A key difference is the amount of common information that all are required to possess.

18. Kobrin 1951. Bob Bursik brought this point to my attention; see Bursik 2009.

19. Pattillo-McCoy 1999; Anderson 1990; St. Jean 2007.

20. Williams and Collins 1995.

21. Raudenbush and Sampson 1999.

22. Thus differentiating our notion of collective efficacy from the aggregation of beliefs about how one’s personal efficacy is dependent on others. See also Bandura 1997.

23. Neighborhood reliability is defined as: Σ [τ₀₀/(τ₀₀ + σ²/nj)] / J, which measures the precision of the estimate, averaged across the set of J neighborhoods, as a function of (1) the sample size (n) in each of the j neighborhood clusters, and (2) the proportion of the total variance that is between neighborhoods (τ₀₀) relative to the amount that is within (α²).

24. Sampson, Raudenbush, and Earls 1997, 922. Analyses were carried out using neighborhood clusters (NCs) and later replicated with tract and community areas.

25. Morenoff, Sampson, and Raudenbush 2001.

26. Browning, Feinberg, and Dietz 2004. To judge whether neighborhood structures serve collective needs, I argue that the theory of collective efficacy must apply the nonexclusivity requirement of a social good—that is, does its consumption by one member of a community diminish the sum available to the community as a whole? Safety from crime is a quintessential social good that yields positive externalities of benefit to all residents of a community, especially its children. So are good schools and clean air. By contrast, we would not consider racial exclusion in the form of racially defended neighborhoods to be a collective good, as for example when neighborhood associations in some American cities in the 1960s and 1970s were exploited by whites to keep blacks from moving to white working-class areas.

27. Small 2009.

28. Morenoff, Sampson, and Raudenbush 2001.

29. Pratt and Cullen 2005.

30. See also reviews in Sampson, Morenoff, and Gannon-Rowley 2002; Kubrin and Weitzer 2003; Leventhal and Brooks-Gunn 2000.

31. Sampson, Morenoff, and Raudenbush 2005.

32. Bernasco and Block 2009.

33. Wikström et al. 2010.

34. Browning, Leventhal, and Brooks-Gunn 2005, 2004. Collective efficacy appears to significantly delay the timing of first intercourse but only for those youth who experience lower levels of parental supervision, suggesting a person-environment interaction.

35. Maimon and Browning 2010.

36. Browning 2002.

37. See Cagney and Browning 2004; Morenoff 2003; Wen, Browning, and Cagney 2003, and Browning and Cagney 2002; and Browning et al. 2006, respectively.

38. Sampson and Wikström 2008.

39. Przeworski and Tuene 1970.

40. Sampson and Wikström 2008, figure 5.3.

41. Sampson and Wikström 2008, table 5.6.

42. For the multivariate model with the structural predictors represented in figure 7.2 controlled simultaneously, the t-ratios of collective efficacy were –4.55 and –3.09 for Chicago and Stockholm, respectively (p < 0.01). Disadvantage was significantly positive in both cities (p < 0.01) followed by minority-group segregation (p < 0.10). I also conducted a multilevel analysis of survey-reported victimization controlling for nine major characteristics of individual respondents in both cities. The odds of violent victimization were higher in neighborhoods characterized by neighborhood instability, disadvantage, and lower collective efficacy, consistent with the picture painted by the independent measures of violence derived from geocoded police records in each city (each of which have unique error sources).

43. Mazerolle, Wickes, and McBroom 2010.

44. Zhang, Messner, and Liu 2007.

45. Kamo et al. 2008.

46. Odgers et al. 2009.

47. Skrabski, Kopp, and Kawachi 2004.

48. Villarreal and Silva 2006.

49. Cerda and Morenoff 2009.

50. http://www.nationalchildrensstudy.gov/Pages/default.aspx.

51. Cohen et al. 2006.

52. Way, Finch, and Cohen 2006.

53. Neighborhood reliability is a function of sample size and the proportion of the total variance that is between neighborhoods relative to that within neighborhoods. The 2002 survey yields lower reliabilities overall, given its substantially lower sample size. For example, at the neighborhood-cluster level the 2002 reliability is only 0.54 compared to about 0.80 in 1995. The corresponding community-area level reliability is 0.92 in 1995 and 0.76 in 2002.

54. For reasons noted above, the correlations are lower but still significantly positive at lower levels of aggregation—0.59 for neighborhood clusters and 0.43 for census tracts (p < 0.01).

55. See also Molotch, Freudenburg, and Paulsen 2000; Suttles 1984. The previous chapter tackles the disorder thesis more explicitly.

56. For an analysis of trust and diversity, see Sampson and Graif 2009a.

57. I thank Richard Block of Loyola for providing the data. I examine census tracts as the main operational unit of variation to gain statistical power and reduce the covariation among predictors. As a check, I repeat key analyses on community areas and block groups. Rather than present a welter of results, I focus on dominant patterns with tracts as the lead case. Because homicide is rare, I analyze the natural log of the rates in five-year intervals—1995–99 is the wave 1 outcome and 2002–6 is the wave 2 outcome.

58. For a description of legal/moral cynicism, see Sampson et al. 2005.

59. Lagged dependent variables are subject to criticism and some feel they induce their own bias by controlling away potential causal pathways. The counterargument is that they account for potential causes of crime that are not explicitly measured. I used the “pretreatment” covariate of lagged homicide in 1993 for wave 1 and 2000 crime for the wave 2 lagged control.

60. I estimated a hierarchical linear model using HLM 6.0. At the temporal level I estimated Crime = β_0j + β_1j CE_ij + Σ _(Controls) β_qX_qij + e_ij, where β_0j is the intercept; CE_ij is the value of collective efficacy associated with time i in neighborhood j; and β₁ is its partial effect on crime. The β_qX_qij coefficients reflect time and other predictors measuring time-varying neighborhood characteristics. The error term, e_ij, is the unique contribution of time, which is assumed to be independently and normally distributed with constant variance α². At the between-neighborhood level the two parameters of interest are β_0j = θ₀₀ + U_0j, and β_ij = θ₀₁, where θ₀₁ reflects the average effect of collective efficacy across j neighborhoods. U_0j is the neighborhood-level error term, assumed to be normally distributed with a variance τ. The variance parameter for collective efficacy was not significant.

61. See Kirk and Papachristos 2011. Although they use a modified scale of legal cynicism that includes indicators of police dissatisfaction, the pattern is clear. Communities with low collective efficacy and high legal cynicism are persistently violent.

62. Coefficient = –0.25, t-ratio = –3.12. The corresponding burglary coefficient was –0.27 (t = –3.15) with lagged burglary rates as a predictor. Because tract-level data provide more cases and induce less overlap among predictors, as another test I entered percentage black as a separate predictor from the concentrated disadvantage scale. In all models disadvantage remained a strong predictor and percent black was positively significant. The estimated effects of collective efficacy were not materially affected, with t-ratios for homicide, robbery, and burglary, adjusted for race, at –3.25, –3.52, and –3.18, respectively (p < 0.01).

63. I estimated a hierarchal model in which years formed the first level and neighborhoods the second. Based on the descriptive trends I examined a quadratic function of time at level 1 (time and time squared, to capture decline and then leveling off), and predicted the slope of the linear part of the crime drop at level 2. In other words, I examined how changes in collective efficacy were related to the magnitude of deceleration in crime, the main component of change. This model controls for time-stable differences between units (heterogeneity of unobserved differences) and examines the trajectory of crime in relation to changes in collective efficacy.

64. Hipp (2009) argues that this applies more to control expectations than cohesion. I agree, but would note that in our formulation “cohesion” refers mainly to trust and perceived helping behavior, not the density of ties, which is measured separately. As Hardin (2002) argues, trust is constituted by shared expectations, so at a theoretical level these concepts are closely related.

CHAPTER EIGHT

1. “Mexicans in Chicago: A New Kind of Politics,” Chicago Tribune, April 6, 2007. These power brokers are members of Chicago-area “immigrant clubs” that are the “engines behind a new political movement that is making itself felt from Illinois to Michoacan.”

2. Obama was director of the Developing Communities Project (Obama 1988, 153).

3. Tocqueville 2000.

4. Putnam 2000.

5. For the initial presentation of the study, see Sampson et al. 2005.

6. For more on this argument, see McAdam et al. 2005.

7. There is a large literature here, but see McAdam 1999 [1982] for a classic source.

8. Wuthnow 1998, 112; Wilson 1987.

9. This thesis harkens back to classic themes in the literature on urban voluntary associations and personal networks as a source of collective capacity (Komarovsky 1946). See the review and more contemporary findings in Fischer 1982 that do not support the declensionist narrative in terms of personal networks.

10. McAdam 2003, 289; Henig 1982.

11. Putnam 2000, 153.

12. Oliver 2001, 202.

13. Alinsky 1946.

14. Marwell (2004, 2007) argues that nonprofit community-based organizations can thus leverage the interests and voting power of their clients. Although Marwell focuses on individual voting, this networking process in principle has the capacity to generate multiple “externalities” of collective civic engagement. A small but important body of recent research supports this idea by showing how nonprofit organizations foster collective action tasks and unanticipated individual gains (Small 2002, 2009; Warren 2004).

15. Earl et al. 2004.

16. Self-help gatherings, unlike a community festival or church pancake breakfast, focus on the individual and are typically not open for public display and consumption. Note that while regularly scheduled meetings (e.g., weekly church services) are excluded, a wide range of nonroutine events that often emerge from the context of regular meetings are included, such as a church fundraiser for AIDS victims, the public claim “PTA Seeks Ouster of Principal,” a school’s fiftieth-year celebration, and special events like “PTA Dads’ Night.”

17. Once event identification was completed, the articles were examined to exclude double counting of the same event. In addition, the articles were scanned to determine if there were multiple events reported in one article. We systematically coded each event within an article based on a coding scheme that was pretested on independent years of data. Interrater reliability was established at both stages of the data-collection process: event identification and event coding. Across persons and sites, interrater reliability averaged 90 percent or more.

18. The distinction between protest and civic events can be captured by noting differences in their forms and claims. Unlike protest, civic events typically do not desire to bring about (or prevent) a change in policy, nor do they express specific grievances. Even though they are also collective in nature, civic events take on different forms than the rally or protest, such as celebrating the community (e.g. festivals, community breakfast), procuring resources (e.g., rummage sales, fundraisers), and accomplishing collective goals (e.g. cleanups, preservation).

19. For a more detailed assessment, see Sampson et al. 2005, 685–87; 703–7.

20. For figures on temporal trends and patterns, see Sampson et al. 2005, 688–91.

21. Taub, Taylor, and Dunham 1987, 184–85; Trice 2008.

22. Total civic memberships correlate highly with civic memberships in one’s residential neighborhood (r = 0.83, p < 0.01) across the seventy-seven communities,suggesting both measures tap the same propensity for civic membership. They also produce similar results.

23. Friend/kinship ties is based on the average number of friends and relatives that respondents reported lived in their neighborhood. Reciprocated exchange is based on a scale of five items: (1) “About how often do you and people in your neighborhood do favors for each other? By favors we mean such things as watching each other’s children, helping with shopping, lending garden or house tools, and other small acts of kindness?”; (2) “How often do you and other people in this neighborhood visit in each other’s homes or on the street?”; (3) “How often do you and people in this neighborhood have parties or other get-togethers where other people in the neighborhood are invited?”; (4) “When a neighbor is not at home, how often do you and other neighbors watch over their property?”; (5) “How often do you and other people in the neighborhood ask each other advice about personal things such as childrearing or job openings?” These items tap the kinds of networks and exchanges traditionally thought to underlie mobilization capacity.

24. As control variables we included predictive (1995) survey measures of community-level socioeconomic status, racial composition (percent non-Hispanic black), and the aggregate level of violent victimization in the community. Population density (persons per kilometer) and population size from the 2000 census are also accounted for when modeling 2000 events. See Sampson et al. 2005 for details on the modeling strategy for collective action events. The results were robust to a number of alternative specifications and sensitivity analyses.

25. Employing lagged outcomes is a stringent test because lagged organizational measures are not available and could have accounted for variation in the rate of 1990 collective action. The lack of statistical significance for later organizational services controlling for lagged collective action thus does not necessarily imply a lack of the former’s causal salience.

26. We were able to identify and geocode the initiating organization in approximately half of all events given the descriptions in the newspaper, compared to an 87 percent location rate for the events themselves. We did not find any information contrary to the assumption that the availability of initiator address information was distributed more or less randomly across areas.

27. Taub, Taylor, and Dunham 1987, 184.

28. This effort is in collaboration with Doug McAdam (see McAdam, Snell-man, and Sampson 2010). The phone fiches list the religious institutions by denomination in addition to telephone numbers and address. We matched the street addresses to a geocoding database maintained by Yahoo and obtained a nine-digit zip code. The addresses that did not match the database were manually checked for spelling errors. Using Google maps, typos and abbreviations were corrected. The result was a total of 11,057 valid church addresses that were matched to census tracts and eventually to community areas in Chicago. Thanks go to Kaisa Snellman at Stanford for research assistance in collection of the NCCS and church data.

29. Indices based on the first principal component of five highly correlated Herfindahl indexes of concentration (see Blau 1977), where Diversity = , and where Π_r refers to a particular group’s (e.g., racial/ethnic) proportion of the population. The constituent measures tapped the number and spread of twenty-five language groups, one hundred birth countries of first-generation immigrants, five race/ethnic groups, and both within-Hispanic and within-Asian diversity. A high score on the summary measure reflects multidimensional diversity in the community.

30. Just less than 60 percent of the variance in total initiator/event location rates was explained. The standardized coefficient for the estimated direct effect of nonprofit organizations per hundred thousand on collective action propensity was 0.39 (t-ratio = 3.93, p < 0.01). Only residential stability had a larger and perhaps surprisingly negative effect. Stable communities appear to be high in informal but not organizational efficacy, once again suggesting that newer forms of collective action do not conform to traditional patterns.

31. Nonprofit density shows a negative relationship with collective efficacy in figure 8.4, but it is not significantly different than zero and therefore directionality is not meaningful.

32. McAdam 1999 [1982]; Cohen and Dawson 1993.

33. Grossman, Keating, and Reiff 2004, 134.

34. For further details on sampling, coding, and long-term trends, see Sampson et al. 2005. Special thanks go to Heather MacIndoe and Simón Weffer-Elizondo for yeoman efforts in training the coders and managing data collection for the CCCP.

35. Because event counts are highly skewed, with a handful of communities like the Loop dominating the overall count and many recording no events, we calculated nonparametric Spearman correlations of rank-order agreement for each decade and for the aggregate total. Each was significant, with the aggregate coefficient (rho) across the three decades at 0.40 (p < 0.01).

36. Sampson et al. 2005, 706. Examples of event overlap include Martin Luther King Day celebrations, Operation PUSH events, fundraisers for UNICEF, a protest over the Contract Buyers League, the Chicago Consular Ball, Hyde Park’s Fourth of July Picnic, and a Lawndale art fair.

37. Obama 1988, 4.

38. McRoberts 2003.

39. This pattern holds across the racial divide. In black communities, the coefficients for the density of nonprofits and churches are 0.30 (p < 0.05) and 0.00, respectively. The corresponding coefficients in nonblack communities are 0.22 and 0.00. The church effect is thus nil, even by race.

40. For a more detailed examination of black civil society and the role of the black church, see McAdam, Snellman, and Sampson 2010.

41. Chicago Tribune, April 9, 2009.

42. Utica, New York, where I grew up, is a shell of its former self, having lost nearly half its population and with abandoned buildings sprawling block after block from downtown. Yet like countless other small industrial cities—many with large minority and poor populations—it is not on the radar screen of the social science or political elite.

43. Putnam, Feldstein, and Cohen 2003.

44. Skogan and Hartnett 1997.

45. Neighborhood poverty may thus be said to have a conditional relationship with organizational participation. See also Swaroop and Morenoff 2006; Small 2004.

46. Fung 2004.

CHAPTER NINE

1. Abraham 2009. The case received widespread attention around the country and beyond.

2. Dispatch: “Were they white, black, or Hispanic?” Lucia Whalen: “Uhm, well, they were two larger men. One looked kind of Hispanic but I’m not really sure. And the other one entered and I didn’t see what he looked like at all. I just saw [them] from a distance and this older woman was worried, thinking someone’s breaking into someone’s house. They’ve been barging in and she interrupted me and that’s when I had noticed, otherwise I probably wouldn’t have noticed it at all, to be honest with you. So I was just calling because she was a concerned neighbor.”

3. “I’ve had much reflection on that, and yes, I would make the call—I would absolutely do that again,” she said. “If you’re a concerned citizen, you should do the right thing.” Boston.com, July 29, 2009, at http://www.boston.com/news/local/breaking_news/2009/07/911_caller_in_g.html.

4. I set aside further discussion of the validity of Gates’s arrest, however, because while important it is not central to my argument.

5. Sugrue 1996.

6. Suttles 1972.

7. Wilson and Taub 2006.

8. Determining whether a crime is motivated by hate, like determining the motive of any behavior, is nonetheless highly problematic. Relying on police accounts renders things even murkier. The finding that neighborhood and countylevel social organization are related to the official reporting of hate crimes (Lyons 2007; McVeigh, Welch, and Bjarnason 2003) suggests the difficulty in separating variations in offending behavior from societal reactions.

9. Rawls [1971] 1999, 461; 1993, 37. To Rawls’s list we can add safety and security. There is a long literature that comes to the same conclusion—across groups there is a great deal of consensus on rankings of seriousness of crimes and the desire for security (Kornhauser 1978).

10. On his communitarian critics, see Taylor 1989; Selznick 1992; Sandel 2009. Perhaps the most penetrating critique, albeit one that is deeply sympathetic, is found in Amartya Sen’s The Idea of Justice (2009). Sen takes issue with Rawls by arguing for a comparative and global framework that emphasizes realized outcomes in addition to idealized (or what Sen calls “transcendental”) institutional principles. In my opinion, however, the critique of Sen does not undermine the basic argument I derive about altruism, much less the essence of Rawls.

11. Rawls [1971] 1999, 458, 463.

12. Gillman (1996) argues that the constitutional law tradition has long sought to balance individual rights against the need to promote the health and safety of communities, and that the very idea of police power signifies the adjudication between individual rights and social order.

13. For a concise explication of the original position, see Rawls 1993, 23–28.

14. For a philosophical analysis that also affirms altruism as a fundamental concept along with prudence, albeit from a different analytical route, see Nagel 1970.

15. Levitt and Dubner (2009) criticize behavioral economic experiments for their artificial setup and maintain that humans can be manipulated to be selfish or altruistic. I agree in part but would revise their critique. Humans respond to social context, and the goal, as in this chapter, is to find naturally occurring instances where behavioral aspects of altruism can be systematically observed (and ideally manipulated) across social settings that vary in theoretically relevant ways.

16. A common experimental game gives players A and B a set amount of money. A can offer any amount to B, and, if accepted, both get to keep the money. If B rejects the offer, no one gets a penny. Rational choice theory predicts that because any amount of money is a gain for both, no offer should be rejected, and A should make a low initial offer. But B will reject A’s offer if it is perceived as unfair, and the opening offers of A are typically much higher than selfish models predict. For recent reviews on social norms, see Stout 2006; more generally, see Thaler and Sunstein 2008. There is also comparative evidence that a city’s rate of helping strangers is a cross-culturally meaningful characteristic of a place (Levine, Norenzayan, and Philbrick 2001).

17. Stout 2006, 14. In terms of political theory, see the discussion by Mansbridge (1990).

18. The main statement on procedural justice is Tyler 1990. The link between Rawls and Tyler’s legitimacy principles deserves further scrutiny. For innovative work on bringing “social identity” into mainstream economic theory, see Akerlof and Kranton 2010.

19. This is the essential idea behind social control theory (Sampson and Laub 1993).

20. For an evolutionary theory of group-based selection on altruism, see Sober and Wilson 1998 and Wilson 2002. On cultural selection and varieties of altruism, see Jencks 1990.

21. Rawls 1993, 16.

22. Sampson and Bartusch 1998.

23. Donations to United Way have been proposed (Chamlin and Cochran 1997). but there are sharp disagreements in society over the role of charity versus the State, and this organization in particular has suffered withering criticism. The financial ability to give is also confounded. Healy (2004) provides an innovative analysis of organizational influences on organ donation.

24. Milgram, Mann, and Hartner 1965, 438.

25. Technically, I used a multivariate logistic response model. For further details and analysis of letter-drop conditions, see also Morenoff and House n.d.

26. In general, results were similar regardless of unit of aggregation.

27. Iwashyna, Christakis, and Becker 1999. I thank Nicholas Christakis for his help in obtaining and interpreting the data.

28. Patterson 2004.

29. I used this fact in additional analysis to create another indicator of CPR by specifying witnessing as an instrumental variable because the proportion of witnessed events varies across communities. Specifically, I used witnessing to predict CPR at the cardiac-arrest level, after netting out the effects of age, race, sex, and home location of victim, and then aggregating by community areas to produce an adjusted CPR rate. This technical analysis is beyond the scope of the present chapter, but the results were nonetheless very similar and give me added confidence.

30. Disadvantage and diversity are defined as in earlier chapters for both 1990 and 2000.

31. Specifically, I created an Empirical Bayes Residual scale that simultaneously adjusts for the cardiac-arrest and neighborhood predictors at the time of event from a multilevel model.

32. In this model, the t-ratio for the organization coefficient was 4.64 (p < 0.01).

33. Specifically, I estimated a multivariate multilevel analysis of 3,303 letters nested within neighborhood clusters and community areas, but the main results focus on the latter. As an outcome I estimate the log-odds of letter returns. Formally, the baseline model can be expressed as:

Letter Model: Prob(Return = 1|B) = P log[P/(1-P)] = B0 + B1*(Post-anthrax) + B2*(Sept. or Oct.) + B3*(Nov.) + B4*(Dec.) + B5*(Jan.) + B6*(Feb. or March) + B7*(Afternoon) + B8*(Windy) + B9*(Rain) + B10*(Address type) + B11*(Observed) + B12*(Weekday) + B13*(Suspicious) + B14*(People on street) + B15*(Residential) + B16*(Parking lot) + B17*(Waterfront) + B18*(High-rise private) + B19*(High-rise public) + B20*(Single-unit public) + B21*(Single private)

Level-2 Model: B0 = G00 + G01*(Size 2000) + G02*(Density 2000) + G03*(Disadvantage 2000) + G04*(Stability 2000) + G05*(Diversity 2000) + U0.

34. The standardized coefficient for concentrated disadvantage was –0.71 (p < 0.01), controlling for all letter-drop conditions and other neighborhood characteristics. Organizational density and population size were also significant predictors (B = 0.34 and 0.26, respectively), whereas race/ethnic diversity, residential stability, and population density were not significant. To explain the disadvantage finding, one person suggested to me that poor neighborhoods might have fewer mailboxes. I have no way of knowing whether this is true, but the research design works against this explanation, in any case. Letters were randomly distributed throughout the neighborhoods, and thus the majority of drops were not in close proximity to a mailbox. One had to pick up the letter and mail it somewhere different from the point of retrieval. I would further argue that the numerous controls for land use and housing type adjusts for “mailbox proximity” across community areas.

35. The reader will not be surprised to learn that blacks and Hispanics are significantly more likely than whites to perceive legal norms as less than binding, and that low SES respondents are twice as likely as high SES respondents are to report high levels of legal cynicism. This pattern occurs in both the 1995 and 2002 surveys. Yet once neighborhood disadvantage is taken into account, blacks hold views about legal norms and police behavior similar to those of whites (see also Sampson and Bartusch 1998). Blacks are more cynical because they are more likely to live in environments of concentrated disadvantage. At the same time, minority groups are more intolerant of deviance and fighting than whites, even taking into account neighborhood-level factors. The analyses in this chapter were rechecked by controlling for neighborhood differences in tolerance of deviance. The results were not affected.

36. The Kaiser-Meyer-Olkin measure of sampling adequacy indicates the proportion of variance in observed variables that is likely caused by an underlying factor. In the present case this figure is 50 percent, which given the elapsed time is surprisingly high and supports, once again, the theoretically based hypothesis of a latent altruistic character.

37. In this analysis, the first principal component accounted for 68 percent of the common variance in letter-return rates in 2002 and collective efficacy 1995–2002. Given the much longer passage of time, it is perhaps expectable that the adjusted CPR rate from 1988 did not correlate strongly with the later survey measure of collective efficacy.

38. Crime rates are measured as the log of the rate per hundred thousand residents for the years 2002 through 2006 for homicide. I also examined burglary and robbery rates. These are widely considered the three most valid and reliably measured of all crimes.

39. For more on this link, see Browning, Leventhal, and Brooks-Gunn 2004, 2005. In addition, I would argue that the assumption of uncorrelated error terms between the altruism scale and teen births is reasonable, given the passage of time and the very different nature of the specific acts in question. Once demographic and economic composition is controlled, the assumption of uncorrelated error terms with respect to crime is reasonable as well.

40. Sampson and Bartusch 1998.

41. Because of the relatively small number of cases overall, I opt for theoretical parsimony in figure 9.2, with control variables limited to the major hypothesized confounders in terms of neighborhood composition. That said, I conducted a number of additional tests to explore the robustness of the main pattern of results. For example, I controlled for alternative demographic markers in the form of concentrated poverty and percent black, along with systemic factors such as residential stability and organizational density. I also examined social process variables such as friend/kinship ties, exchange networks, and organizational density, which I showed above was a key predictor of altruism. Although there were some fluctuations depending on the model specification, the basic pattern remained. The results were largely similar for burglary rates and robbery rates in 2002–6 as well, suggesting the crime pattern is especially robust.

42. For homicide, the standardized coefficients for letter drop/CE and CPR were –039 (t-ratio = –3.80; p < 0.01) and –0.11 (t-ratio = –1.93; p < 0.10), respectively. The comparables for teen birth rates were –0.23 (t-ratio = –2.50) and –0.20 (t-ratio = –3.78), both p < 0.05. The separate indicator of CPR altruism thus has relatively greater predictive power for teen birth rates than homicide.

43. This pattern broadly supports the arguments of Patterson (2004, 2009). It is worth noting as well that I found no evidence that collective efficacy (or trust or cohesion) was associated with selfish norms (so called negative social capital) or anti-altruistic behavior.

CHAPTER TEN

1. Tobler 1970:236.

2. For a debate on the “first law,” see Goodchild 2004 and response in Tobler 2004.

3. For what is probably the classic statement on the spatial ecology of interaction, one still alive and well, see Festinger, Schachter, and Black 1950. More recently, see Blau 1977.

4. I am indebted to Jeff Morenoff and Corina Graif for their collaborative work on spatial analyses in earlier papers. This chapter is not an appropriate forum for details on statistical modeling. See Morenoff et al. 2001 and Graif and Sampson 2009 for more on methodology. For a discussion of spatial models, see also Messner et al. 1999; Tita and Cohen 2004.

5. For a primer on spatial statistics, see Anselin (1988). Moran’s I is the unstandardized regression-slope coefficient of y predicting Wy. Like a correlation, it has an upper limit of 1 so that the higher the number, the higher the positive spatial correlation. A “pseudo”-significance test of spatial correlation is calculated by comparing observed correlations with a distribution of randomly permuted ones. Results are similar with weights based on the four closest neighborhoods.

6. This pattern suggests that the spatial interdependence estimates are conservative, although it should be noted that the density of nonprofit organizations is well measured and yet not very interdependent with its spatial neighbors. As will be shown in chapter 14, organizational mechanisms take on a higher-order structure that goes beyond nearest neighbors.

7. Morenoff, Sampson, and Raudenbush 2001. See also Sampson, Morenoff, and Earls 1999 for a spatial analysis of child-related dimensions of collective efficacy and “social capital.”

8. In earlier chapters I therefore opted to present the simpler rather than more complex findings when the substantive results were robust to the introduction of spatial errors. This chapter is mainly focused on the “value added” of a direct spatial analysis, especially the roles of spatial lag and spatial heterogeneity, to which I now turn, and the concept of spatial advantage.

9. In technical terms we can define y_i as the homicide rate of Neighborhood (N) i, and w_ij as element i,j of a spatial weights matrix that expresses the geographical proximity of N_i to N_j (Anselin 1988, 11). For a given observation i, a spatial lag Σ_iw_ijy_j is the weighted average of homicide in neighboring locations. This can be defined in different ways, but in the present data the patterns are robust to alternative spatial weights. For the most part PHDCN work has used what is called first-order contiguity, which defines neighbors as those Ns that share a common border (referred to as the rook criterion). Thus, w_ij = 1 if i and j are contiguous, 0 if not. Similar results obtain when using the four or six nearest neighborhoods to a focal neighborhood, as in the Moran’s I plots. To test formally for the independent role of spatial dependence in a multivariate model, the spatial lag is treated as an explanatory variable. The spatial lag regression model is thus: y = ρWY +Xβ + ε, where y is an N by 1 vector of observations on the dependent variable; Wy is an N by 1 vector composed of elements Σ_iw_ij y_j, the spatial lags for the dependent variable; ρ is the spatial autoregressive coefficient; X is an N by K matrix of exogenous explanatory variables with an associated K by 1 vector of regression coefficients β; and ε is an N by 1 vector of normally distributed random error terms, with means 0 and constant variances.

10. Technically, the notion of diffusion implies a process that occurs over time, whereas the spatial autocorrelation we have examined are point-in-time— homicide rates are spatially interrelated across neighborhoods and simultaneously determined. Further details about the maximum likelihood estimation are available in published work (Morenoff, Sampson, and Raudenbush 2001).

11. Although spatial error cannot be definitively distinguished from spatial lags, there are empirical tests to explore which process is more consistent with the data. I conducted these tests and only estimated spatial lag models where the data indicated a better fit and where there was strong theoretical motivation to posit spatial diffusion or exposure-like effects. The fact remains, however, that the assumptions behind a spatial lag model are not possible to prove in the data.

12. Morenoff and Sampson 1997.

13. Sampson, Morenoff and Earls, 1999.

14. Morenoff, Sampson, and Raudenbush 2001.

15. Sampson and Wilson 1995.

16. African Americans are approximately six times more likely to be murdered than whites (Fox and Zawitz 2003), and homicide remains the leading cause of death among young African Americans (Anderson 2002). That alone makes violence surface as perhaps the major challenge to the viability of cities and pursuit of racial equality. Both police records and self-reported surveys continue to show disproportionate involvement in serious violence among blacks, and nearly one in three black males will enter prison during their lifetime compared to less than 5 percent of white males. Even as crime continues to decline, African Americans are at increasing risk of incarceration and subsequent weak attachment to the labor force (Western 2006), reinforcing black disadvantage and involvement in crime.

17. This conclusion is confirmed in two assessments of the available literature from 1995 to approximately 2005 (Peterson and Krivo 2005; Pratt and Cullen 2005). In the meta-analysis of Pratt and Cullen, concentrated neighborhood disadvantage is the largest and most consistent predictor of violence across studies. Peterson and Krivo find substantial similarity in the basic pattern of relationship with both blacks and whites. There are exceptions here and there, as expected, but the general trend is thus more similarity than dissimilarity in basic pattern.

18. Shaw and McKay’s ([1942] 1969) observation in Chicago from over half a century ago is worth repeating: “The important fact about rates of delinquents for Negro boys is that they too, vary by type of area. They are higher than the rates for white boys, but it cannot be said that they are higher than rates for white boys in comparable areas, since it is impossible to reproduce in white communities the circumstances under which Negro children live. Even if it were possible to parallel the low economic status and the inadequacy of institutions in the white community, it would not be possible to reproduce the effects of segregation and the barriers to upward mobility” (614).

19. Pattillo-McCoy 1999.

20. Sampson, Morenoff and Earls (1999) constructed hypothetical simulations in which the mean level of Wy in white neighborhoods was assigned to black neighborhoods, thus equalizing the spatial inequalities between the two groups. The results suggested that, all else being equal, giving black neighborhoods the same mean spatial proximity scores as white neighborhoods reduces the racial gap in child-centered social control by 38 percent and the gap in adult-child exchange by 64 percent. Performing the same exercise using concentrated disadvantage rather than spatial proximity reduces the racial gap in child control by 56 percent, but produces no change in the racial gap in adult-child exchange. These simulations provide further evidence that the differing spatial environments of black and white neighborhoods play a role equal to if not greater than that of internal structural characteristics in generating inequalities in collective efficacy.

21. Saenz 2004.

22. Martinez 2002; McNulty and Bellair 2003a, 2003b.

23. Sampson, Morenoff, and Raudenbush 2005.

24. The estimated probability that an average male living in a high-risk neighborhood without immigrants will engage in violence is almost 25 percent higher than in the high-risk immigrant neighborhood, a pattern again suggesting the protective rather than crime-generating influence of immigrant concentration. This finding is broadly consistent with other research (Martinez 2002).

25. Sampson 2006, 2008c.

26. Sutherland 1947.

27. We first constructed a Herfindahl index of the concentration in census tracts of twenty-five distinct foreign languages in order to capture a broad range of immigrant-related diversity. A second measure of percent foreign born yielded a broadly similar albeit somewhat weaker pattern. For details, see Graif and Sampson 2009, from which Figure 10.3 is adapted.

28. Interestingly, the estimated effects of collective efficacy remained significant but did not vary spatially, suggesting a space-invariant process unlike immigrant diffusion.

29. Anderson 1999.

30. Huntington 2004.

31. Nisbett and Cohen 1996.

32. For more on my immigration penetration thesis, see Sampson 2008c.

CHAPTER ELEVEN

1. Briggs, Popkin, and Goering 2010.

2. See “American Murder Mystery” (Rosin 2008).

3. Kling, Liebman, and Katz 2007, 109. See also Oakes 2004. Considering that over a century of neighborhood ecological research forms the baseline, this is high praise for one study.

4. NBER 2006.

5. Clampet-Lundquist and Massey 2008. Destination neighborhoods were still segregated (> 75 percent black on average), whether for the control or experimental group. See also Sampson 2008b.

6. Note that situational effects can still be persistent and form an enduring pattern. For example, our research in Chicago shows that while collective efficacy consistently predicts the event rate of violence in a neighborhood, including into the future (chapter 7), it does not predict rates of offending by neighborhood youth, the latter of which may occur anywhere (Sampson, Morenoff, and Raudenbush 2005).

7. Schlossman and Sedlack (1983) provide a review of the idea and empirical content of the community-level interventions animating the Chicago Area Project (CAP) up until the 1980s, which grew out of the delinquency-related research of Shaw and McKay reviewed in chapter 2. CAP celebrated its seventy-fifth anniversary in 2010. Population-level interventions are better established in public health, including the experimental randomization of macrolevel ecological units to causal treatments (Boruch and Foley 2000; Sikkema et al. 2000).

8. The HOPE VI federal housing program and the New Communities Program of the MacArthur Foundation (http://www.newcommunities.org/) represent two examples of a governmental and private intervention at the macro level, respectively. I make no claims here on success or failure, the point is to further contrast the individual versus neighborhood question. For a counterfactual approach to neighborhood interventions, see Verbitsky and Raudenbush 2006.

9. Orr et al. 2003, appendix C2.

10. Langer and Michael 1963.

11. Although for somewhat different reasons, the indiscriminate use of control variables was the subject of a detailed warning by Stanley Lieberson over twenty years ago (1985, chap. 6).

12. See Clampet-Lundquist and Massey 2008 and Ludwig et al. 2008.

13. Sampson 2008b. This chapter refines my earlier argument and presents new data.

14. Following MTO publications, I begin with census tracts as analytic units because they permit me to link geocoded address data over time. For the city of Chicago and the United States as a whole (representing over sixty-five thousand census tracts), one linear factor was extracted on the basis of a principal components analysis of poverty, welfare assistance, female-headed families, unemployment, and percent black (see also Sampson, Sharkey, and Raudenbush 2008). I thus focus on a summary scale of concentrated disadvantage rather than a single item. In later analysis I look at measures beyond disadvantage.

15. Sampson 2008b, figure 1.

16. The racial composition difference was not significantly different across treatment groups. It is important to note that “scaling up” to account for compiler status simply scales up standard errors too, and thus does not change the significance (or not) of differences.

17. Linear differences in the principal components scale of concentrated disadvantage, which weights poverty more than race, are significant by treatment group (p < 0.05). A z-score scale including all U.S. and Chicago area moves is not significantly different.

18. All t-tests use randomization weights recommended in Ludwig et al 2008. As noted above, adjusting for compliance status does not change any conclusions regarding significance.

19. What about interim moves? I have focused on destination neighborhoods for theoretical reasons, but consistent with MTO publications I found a significant (p < 0.01) nine-point difference between the experimental and control group in duration-weighted poverty. However, the duration-weighted difference for percent black is only two points and not significant (p < 0.05).

20. This is consistent with prior research showing no effect on neighborhood employment change (Kling, Liebman, and Katz 2007, web appendix, table F14).

21. Residents in the 2002 community survey were asked: “How many close friends and relatives do you have (people that you feel at ease with, can talk to about private matters, and can call on for help)?” “How many friends and relatives do you have to whom you can turn when you need to borrow something like a household object or a small amount of money or need help with an errand?” “How many friends and relatives do you have who you can ask for advice or information?” In each question respondents provided the number of friends. I created an overall scale based on the mean for each Chicago community across the three dimensions.

22. These indicators were highly interrelated and comprised just one factor, so I took the first principal component to reflect the common variance.

23. I selected a random sample of the experimental group to equalize sample sizes and provide a balanced and easier-to-see comparison. Using the full sample produces a similar pattern.

24. See also Sampson 2008b, figure 5.

25. There were no significant differences between treatment and control groups at the community-area level for the eleven survey-based social processes independently measured in PHDCN, both level and change. I also examined spatial “neighbors of the neighborhood” at the census-tract level by calculating the spatial lag of poverty and disadvantage using a Euclidean distance-based measure—the average of neighboring tracts weighted by distance from the focal tract. Neither spatially lagged poverty nor concentrated disadvantage were significantly different between treatment and controls. At destination the magnitude of difference in spatially lagged percent poverty was less than 1.5 percentage points.

26. For a disadvantage map, see Sampson 2008b, 220.

27. For maps of Baltimore moves in MTO, see Clark 2005.

28. Sharkey 2008. According to Sharkey, this means that in a fundamental sense the “ghetto is inherited.” He is further examining this notion in a book in progress.

29. Zorbaugh 1929, 134.

30. Basically, SUTVA means that potential outcomes for one unit are unaffected by the particular assignment of treatments to other units. For further discussion, see Sobel 2006.

31. Deaton 2008, 4.

32. Wheaton and Clarke 2003.

33. Shonkoff and Phillips 2000, chap. 8.

34. Sharkey 2008.

35. Sampson, Sharkey, and Raudenbush 2008.

36. I set aside the statistical details here, but the main strategy was to adapt a method pioneered by Jamie Robins for causal inference with time-varying data: inverse probability of treatment weighting (IPTW); see Robins 1999; Robins, Hernan, and Brumback 2000; Hong and Raudenbush 2008. The intuition is that one wants to give more priority to data observations that are unconfounded with observable predictors of selecting into the treatment. In the following chapter I will present validating information on the causes of residential sorting that were used in the model. Because whites and Hispanics were not exposed to concentrated disadvantage in anything near a comparable way, our analysis of verbal ability focused on African American children. For details see Sampson, Sharkey, and Raudenbush 2008.

37. Shonkoff and Phillips 2000.

38. See also Sampson 2008b. The t-ratio for this estimated effect is 2.74 (p < 0.01). When including twelve-year-olds there is also a significant cohort-by-treatment interaction.

39. Sampson 2008b, 852.

40. Informative models of selection and neighborhood effects can still be applied to MTO. Corina Graif is exploring spatial effects in her Ph.D. dissertation in all five MTO sites, and long-term follow-up of the MTO will allow for a more rigorous test of developmental effects (see also Ludwig et al. 2008). To the extent that significant (although still marginal) differences in the local neighborhood and larger spatial environment are maintained as a result of the original voucher offers, then the chances of observing developmental treatment effects will be increased. Moreover, a recent study comparing the Chicago samples of MTO and PHDCN along with another voucher study found similar negative effects of concentrated disadvantage on verbal ability (Burdick-Will et al. 2011). This relationship thus appears to be sturdy across studies and method of assessment.

41. This section updates my earlier assessment in Sampson 2008b. I return to a discussion of experimental design and social causality in the final section of this book.

42. By following the youngest MTO children further in time, one can gain more leverage on developmental interactions, making the third follow-up of MTO an important scientific investment even though the design limitations analyzed in this chapter will remain.

43. Technically, SUTVA means that the causal effect estimate in MTO is the difference between the average treatment effect and the “spillover” effect on the untreated (Sobel 2006, 1405). Not only are spillover effects of great substantive interest, policy inferences could be led significantly astray by not distinguishing these two components of the treatment effect.

44. Hudgens and Halloran 2008; Rosenbaum 2007; Berk 2005. I set aside here discussion of other and perhaps even more demanding assumptions that analysts of the MTO must adopt in order to make causal inferences, such as the assumption that compilers in the experiment were no more likely to benefit from a neighborhood move than others. This is a strong assumption, as is the assumption of parallel effects of posttreatment mediators. None of these assumptions are subject to direct empirical verification and thus must be asserted as part of a social theory.

CHAPTER TWELVE

1. Ahmed and Little 2008.

2. Ludwig et al. 2008, 150.

3. Heckman 2005.

4. Alba and Logan 1993. See also Logan and Alba 1993; Logan et al. 1996.

5. Massey and Denton 1985.

6. Logan 1978. But see chapter 2, where I noted the nuances in Chicago School theory that, with the exception of Suttles (1972), have gone largely unremarked.

7. See, e.g., Massey, Gross, and Shibuya 1994; South and Crowder 1997,1998.

8. Crowder, South and Chavez, 2006.

9. This chapter builds on the analysis in Sampson and Sharkey 2008. A later review of selection bias and residential mobility consistent with this chapter is Bergstrom and Ham 2011.

10. Faris and Dunham 1939 [1965].

11. Clausen 1991.

12. Measure based on the revised conflicts scale (Straus et al. 1996), with reliability 0.84.

13. Walters et al. n.d.; Kessler and Mroczek 1997.

14. Elder, Johnson, and Crosnoe 2003; Settersten and Andersson 2002.

15. For descriptive statistics on neighborhood outcomes and all covariates at the individual and family levels, along with details on the analytic modeling strategy, see Sampson and Sharkey 2008. If individuals dropped out of the study, their information was excluded only in the wave(s) in which they do not take part in the survey. We adjusted for any bias that might arise due to attrition by modeling the probability that individuals would leave the study at each wave, and then weighting the data in the time-varying file based on the inverse probability of attrition. Also, for individuals who were interviewed in a given wave but did not provide information for a specific variable, we used a multiple imputation procedure for missing values. Finally, to estimate the effects of within-household change (e.g., in income) on change in neighborhood environments, predictors of neighborhood attainment are decomposed into between-individual (time-stable) and within-individual (time-varying) components. Approximately a quarter of the variance in neighborhood income is due to change over time.

16. The reliability of the slope of change in median income for stayers in the neighborhood is 0.59; for movers out of the city, it is 0.84; and for movers within the city, it is 0.75. These results confirm the substantial “signal” in the data in terms of differences in the rate of change in neighborhood income over time. All results are weighted to account for potential attrition bias.

17. Results were nonetheless consistent with expectations. For example, consistent with the spatial assimilation model, household income and education are strong predictors of neighborhood income. Compared to high school graduates, caregivers with at least a college degree are estimated to live in neighborhoods where the median income is roughly $2,500 higher. Compared to families with total household income in the range of $30,000-$40,000 per year, being in the highest income group (household income over $50,000) is associated with a jump in neighborhood median income of about $7,400. By contrast, being in the lowest income group (less than $10,000) is associated with a drop in neighborhood median income of $2,500 relative to the reference group. Change in household income leads to improvement in the neighborhood environment—individuals who entered the highest income group during the course of the study shifted their residences into more affluent neighborhoods. See Sampson and Sharkey 2008, 15–19.

18. Rossi 1980. On later work, see Lee, Oropesa, and Kanan 1994; Crowder 2000.

19. Sampson and Sharkey 2008, table 4.

20. On racial attitudes, see Charles 2000. For an analysis that provides direct evidence on the reemergence of racial segregation among those who moved to integrated neighborhoods, see Sharkey 2011.

21. T-ratio = 2.15 (p < 0.05). I also simultaneously examined measures of other social processes, such as exchange networks, personal victimization, and organizational participation in the neighborhood, but none of these were significant. Shared perceptions of disorder stand out.

22. Herrnstein and Murray 1994; Wilson and Herrnstein 1985. Although IQ and impulsivity measures were only available for children in the household, the “nature” argument rests on the reported high degree of heritability in both constructs across generations. I took the means for those families with two or more children.

23. Kirk 2009a.

24. Details presented in Sharkey and Sampson 2010. This analysis takes into account and adjusts for all the stable and time-varying confounders described earlier in this chapter. Using a dynamic modeling strategy described in Sampson, Sharkey, and Raudenbush 2008, our basic strategy was to first model the predictors of moving and then use these results to inform an analysis that isolated the time-varying effects of moving on later individual outcomes. Analogous to a propensity score method, we obtained balance on the covariate set and describe the pros and cons of assumptions underlying the method. We also used an “instrumental variable” approach that makes different assumptions but yields similar results.

25. There was no significant effect of moving out of Chicago on violent victimization for the propensity-based method, but there was using an instrumental variables approach. Interestingly, further analysis showed that moving was largely unrelated to other nonviolent outcomes, such as school performance, alcohol use, and verbal ability. The specificity of the findings suggests that a method artifact is not driving the consistency of the violence results.

26. With the exception of Quillian’s (1999) analysis, studies of change in neighborhood attainment typically focus on individual processes and not their aggregate consequences. There is a rich literature that uses formal simulations to estimate how neighborhood racial segregation is produced. Schelling (1971) is the classic statement on how individual preferences translate into a segregated residential structure. Through simulations, he showed that collective sequences of mobility decisions made by individual actors responding to their divergent preferences for neighborhood composition can lead to more segregation than any individual would prefer, a “threshold” or tipping-point effect. However, using data on neighborhood preferences, Bruch and Mare (2006) find that the probability of moving increases in a continuous rather than nonlinear manner as more and more members of an outgroup enter the neighborhood. The lower levels of segregation produced by Bruch and Mare’s simulations raise the possibility that preferences, by themselves, cannot explain residential patterns in America.

27. Similar to any research design that follows the same individuals as they move over time, we cannot capture changes in neighborhood composition due to inmigration by nonsample members (e.g., migration from Mexico or into Chicago from the suburbs). The analysis of flows is also constrained by the supply of destination neighborhoods, especially nonpoor areas in the suburbs with predominantly minority populations.

28. For details see Sampson and Sharkey 2008, 24. Rather than analyze the difference between the number of a certain race/income group making a transition in one direction and the number making the same transition in the opposite direction (e.g., Quillian 1999), we examined flows between neighborhoods occurring in each direction separately. We chose the threshold of a 5 percent transition to balance a concern for discovery with a concern for emphasizing common, systematic patterns of movement across neighborhoods. In other analyses we explored race-speciflc flows, different thresholds for transition, alternative definitions of poverty, and applied design stratification weights. The basic flow patterns were robust.

29. This flow consists almost entirely of Latinos moving from predominantly white origin areas into Latino neighborhoods. Race-speciflc flows also document that 80 percent of whites transition into (or remain in) neighborhoods that are predominantly white and nonpoor.

30. The same holds for other communities that have achieved stable integration: “In places like Beverly. . , there are institutions to promote inclusiveness and integration” (Trice 2008).

CHAPTER THIRTEEN

1. Christakis and Fowler 2009; Watts 2003. For a review of the major insights of network research in the social sciences, see Borgatti et al. 2009.

2. See Fischer 1982.

3. Graif and Sampson 2010; Sampson and Graif 2010.

4. Massey and colleagues (1990) present a leading example of international migration as a network.

5. Household address information at all three waves of the PHDCN was geo-coded and linked to census-tract codes, allowing us to trace changes in the residential locations of sample members occurring over the course of the study by tracts, neighborhood clusters, and community areas. Ties (arrows) are adjusted to account for sampling weights at the household level, and moves are calculated as a proportion of the sampled neighborhood population at baseline. Sampling weights account for the stratified neighborhood selection and the within-cluster probabilities of selection based on age of child. Weighted data are thus reflective of the targeted Chicago population. Moving flows outside the city are not shown, but analyses below adjust for the differential probabilities of residents in each neighborhood to leave the city. Similar results occur for indegree and outdegree based on the eighty sampled neighborhood clusters (Graif and Sampson 2010: Sampson and Graif 2010).

6. Any community in Chicago can receive movers and so indegree reflects all areas. Over 85 percent of Chicago neighborhoods received a PHDCN mover. By contrast, neighborhoods that were not included in the original sample have an outdegree (and corresponding size) of zero.

7. The reader can thus zoom in on any community of interest. To link names to area identifiers not named in the figure, see Chicago Fact Book Consortium 1990 or http://www.cityofchicago.org/city/en/depts/doit/supp_info/citywide_maps.html.

8. McPherson, Smith-Lovin, and Cook 2001, 431. See also Blau 1977.

9. Wellman 1996.

10. Burt 1987; Galaskiewicz and Wasserman 1989.

11. Price-Spratlen 1999; Massey et al. 1990.

12. Sugrue 1996; Bursik and Grasmick 1993.

13. The thesis of the defended community would also predict this pattern (Suttles 1972).

14. Details for the underlying analyses are elaborated in Graif and Sampson 2010. We applied a multiple regression quadratic assignment procedure (MRQAP) that excludes the diagonals (Krackhardt 1988, 1992) and accounts for network-specific dependencies in the dyadic neighborhood data. The data are randomly permuted to form a reference distribution under the null hypothesis of no relationship, from which parameter estimates are calculated and compared against observed estimates to assess significance. Because of this nontraditional method, I emphasize coefficients significant at the 0.05 level. The analysis explicitly controls for dissimilarity between neighborhoods in the proportion of original residents that moved out of Chicago and dissimilarities in both indegree and outdegree defined earlier.

15. Tests for the overall model indicate that the proportion of random trials yielding a value as large as or larger than the observed estimate of explanatory power is lower than 1 in 1,000. There was only a 3 in 1,000 odds of observing the spatial distance coefficient by chance.

16. Because the sample size of individuals is smaller in nonsampled clusters, this analysis examined binary ties defined by the presence of any move in any direction. The full city analyses also introduced a sample design control, defined as a matrix whose cells equal unity if neighborhoods within dyads were both part of the initial sample, and zero otherwise.

17. This includes the sampled network and all Chicago NCs, with results significant in each household income category except for the high income, sampled network cell (p < 0.10).

18. Kefalas 2003; Carr 2006.

19. Pattillo 2007; St. Jean 2007.

20. See the rich account in Kefalas 2003 on the importance of perceived orderliness and the shared narrative of the “The Last Garden” in one of the Southwest Side communities encircled in the right panel of figure 13.4.

21. Pettit and Western 2004; Sampson and Laub 1993; Western 2006.

22. For the full PHDCN sample, I created an indicator based on domestic violence (self-reported by caretakers) and whether any family member had an arrest record. I define “criminal” households as those with unofficial violence within the family and a family member with an official arrest history. I also examined a more restrictive measure of adolescent violence based on a self-reported scale validated elsewhere (Sampson, Morenoff, and Raudenbush 2005), with the violent group defined as the highest quartile. The results are similar, but this measure is only available for those PHDCN subjects over twelve years of age, so I focus on family criminality.

23. With few exceptions, these results hold independent of the definition of ties (weighted or binary) and the network analyzed, whether based on the sampled nodes or all the nodes.

24. Sample sizes are not large enough to reliably examine mobility flows by IQ or impulsivity, as these were only measured on children in selected cohorts (see chapter 12).

25. The probability of moving out of Chicago was invariant by depression like it was for crime, and therefore limiting to the city does not bias the results. Fifteen percent of depressed caregivers moved out of the city compared to 16 percent of nondepressed caregivers (not significant). The main pattern of results was also robust to the type of tie—for example, volume of flows or any connection—and to the network analyzed (sampled NCs or citywide).

CHAPTER FOURTEEN

1. Laumann and colleagues in particular made important contributions to the study of elite influence in a small German town and in two American communities (Laumann and Pappi 1976; Laumann, Marsden, and Galaskiewicz 1977). See also Galaskiewicz 1979 and Knoke 1990.

2. Gould 1991 represents an important exception to this general observation. Perhaps the most direct parallel to the current study at the level of between-community comparison of networks isthe Rang Nong Study of fifty-one village networks in Thailand (Entwisle et al. 2007). More generally, the ADDHEALTH study is a recent innovation that has improved our ability to study networks and social context.

3. Janowitz 1975; see especially the critique in Marwell 2007.

4. Campbell 1955.

5. Houston and Sudman 1975; Laumann and Pappi 1976.

6. For example, 93 percent of KeyNet respondents corroborated the community-area name, and 84 percent reported using its officially designated boundaries in carrying out their work.

7. Burt 1995.

8. Granovetter 1978, 1541.

9. An examination of individual ties in selected communities where I had detailed knowledge of actual connections (e.g., Hyde Park) suggested a high degree of construct validity. The leaders at the center of the figure or in the center of cliques are also well known in Chicago. In follow-up qualitative work I interviewed a set of these leaders, described the KeyNet study, and probed the nature of individual and institutional connections. This project is beyond the scope of the present chapter and will be separately reported.

10. For details on network measures generally, see Burt 1980; Wasserman and Faust 1994. For further discussion of the application to the KeyNet study, see Sampson and Graif 2009a.

11. Granovetter 1973, 1374; see also Gans 1962.

12. But as Gans (1974) responded, it was not that bridging ties within the West End were absent, rather it was that external ties to the city’s political elites were missing. Either way the network ties went unexamined empirically by both sides in the debate.

13. Because we allowed up to five nominations, the denominator is (5*n*(n-1)), where n refers to the number of citations.

14. For each exponent k, the number in a cell indicates the total number of k-step ties that connect respondent i and respondent j (Burt 1980, 86–88). A new matrix was computed in which the cells register the minimum number of steps it takes for respondent i to reach respondent j. In no case did more than four steps add to the connectedness of these networks, so I stopped at four. I then recoded all nonzero equal to 1 and computed a new density measure: total # of direct or indirect ties divided by n(n-1). I thank Ezra Zuckerman for early assistance in creating these measures for the original study and Dave Kirk and Corina Graif for research assistance, especially in the second KeyNet panel.

15. Note that community of home residence is a separate matter. One can be a principal in Logan Square, for example, and live in Lincoln Park. For purposes of the analyses in this chapter, I define internal and external ties not on the home residence of the leader but the community in which he or she works or has jurisdiction over.

16. More formally, Σ_j [(Σ_iX_ij) /#NOM]², where #NOM=Σ_iΣ_jX_ij and i indicates a respondent from community k, j refers to an alter nominated by respondent i in community k, and X_ij represents a tie (or nomination) from respondent i in community k to any alter j of all alters receiving nomination from respondents in community k. X_ij equals 1 if a tie exists and 0 if it does not.

17. Putnam 2000.

18. Let p_bc(a) be the proportion of all shortest paths linking community b and c which pass through a. The “betweenness centrality” of community a is defined as the sum of all p_bc(a) across all pairs of communities, not including a, where a, b, and c are distinct communities in the citywide network of dichotomized ties and the sum is adjusted for network size. Betweenness centrality is a function of the number of times a given community serves as the shortest pathway between any other two communities (Freeman 1977), hence the idea of centrality.

19. Network analysis raises complex issues of bounding and what one defines as the full network. Based on the sampling design, I argue that the KeyNet study either saturated or nearly saturated each domain and that in the latter case the selected respondents are a representative sample. Given the multistage probability sampling of PHDCN areas to begin with, I further argue that the network variability is comparable across the sampled network of forty-seven community areas (and thirty areas in 2002) and that these areas are reflective of nonsampled communities in Chicago. Another analytic issue that is sometimes raised is the size of the network, as there may be an inherent tendency for larger networks to show less density. Although a particular network measure can be sensitive to design features, I tried to err on the conservative side by examining multiple measures wherever possible and relying on those with evidence of construct validity, a relatively simple interpretation, and that produced similar results despite variations in measurement or model specification and adjustment for network size.

20. Community areas are arrayed geographically but not with the precision of typical GIS maps. The network software positions nodes to accommodate size as well (in this case outdegree), so that in some cases the circles are not in the exact geographic center of a given area.

21. Specifically a cohesive leadership style as indicated by the higher centralization of elite contacts predicts lower homicide rates in a community (t-ratio = –3.09), second only to concentrated disadvantage. Internal centralization yields a similar but slightly stronger link to lower violence. Cohesive leadership predicts lower teen births as well, but at a reduced level of significance (p < 0.10), after concentrated disadvantage and organizational density. By contrast, low birth weight is unrelated to leadership networks once these other characteristics are controlled.

22. Network analysis is frequently criticized for being overly technical and emphasizing measurement at the expense of substance, producing highly refined analysis on networks for which there is little inherent interest. I have tried to steer away from a focus on the technical and have emphasized instead the major patterns in the data for theoretically motivated constructs. But one could certainly go further, such as defining subgroup clusters (block modeling) and more refined embeddedness measures for both communities and organizations. I set this task aside for future work.

CHAPTER FIFTEEN

1. Even the discipline of economics, long thought to be hostile to the idea of social context in general and neighborhood effects in particular, is radically changing in ways that I believe this book furthers (see especially the effort of Akerlof and Kranton 2010).

2. Glaeser 2011.

3. Raudenbush and Sampson 1999, chap. 7.

4. http://www.census.gov/apsd/techdoc/cps/cpsnov08c.pdf (see attachments 7 and 8).

5. The National Children’s Study (NCS) in the U.S. is also likely to include ecometric measures of contextual social processes. At the societal level, see Hall and Lamont 2009.

6. See, for example, Wikström et al. 2010.

7. Go to http://worldmap.harvard.edu/chicago.

8. See Galster 2011 for a similar argument on the importance of analyzing neighborhood effects at multiple scales of geography. The work of Grannis (2009) lays out the idea of tertiary communities built up by microlevel social interactions that are shaped by street layout.

9. Mumford 1954, 258.

10. Smith 2010,1

11. Patterson 2004, 2009.

12. Note that such a critique is teleological and subsumed in a macrostructural determinism. That durable inequalities persist across large-scale political and economic change undermines the idea that we can reduce such continuities to only outside forces.

13. Cooley 1902.

14. Suttles 1984; Permentier, Hamm, and Bolt 2009.

15. Chapter 7; Sampson and Bartusch 1998; Sampson, Morenoff, and Raudenbush 2005; Kirk and Papachristos 2011.

16. Patterson 2009, 1. I make no pretense here to be comprehensive in evaluating culture writ large or the seminal work of Bourdieu on cultural reproduction and “habitus” (Bourdieu 1977). On the post-Bourdieu “cultural turn,” see Lamont 2000; Small, Harding, and Lamont 2010; Patterson 2004, 2009; Harding 2010. See also Wilson 2009 on the integration of structural and cultural forces within the context of explaining inner-city life, and Small’s (2004) integration of Goffman’s (1974) frame analysis with neighborhood narratives on organizational participation. Economists have also begun to take culture seriously (Akerlof and Kranton 2010; Akerlof 1980) and more recently a pushback has emerged among sociologists to reclaim culture as causal motivation rather than just rationalization, a position I endorse (see especially the theoretical and empirical project of Vaisey 2009). My focus here is on neighborhood processes and the explanation of continuity.

17. The apparent rejection of any talk of values or norms by the new cultural sociology is rooted, by my reading, in a “straw-man” version of Parsons. While Parsons certainly overreached, to explain deep continuities in social life without recourse to the role of social norms seems shortsighted and, in the end, unconvincing.

18. Mazerolle, Wickes, and McBroom 2010.

19. See also the analysis in Sampson and Graif 2009.

20. Hampton 2010.

21. Collective efficacy at the school level is studied by Kirk 2009b; see also Bryk and Schneider 2002. It does not seem a stretch to imagine that the theoretical ideas of collective efficacy presented here could even be applied to financial firms and the recent economic crisis, which I would argue stems from a failure of control (social regulation) combined with a weakening of shared expectations about moral behavior (e.g., a cynical maximizing of short-term profit for personal gain at the expense of the long-term wellbeing of a company). Cultural norms and deterrence processes might go a long way toward explaining Wall Street actions.

22. This may explain why, in some societal contexts, the cohesion component of collective efficacy appears less important or even a separate factor. In the favelas of Brazil, for example, historical forces have concentrated forms of severe disadvantage such that cohesion may be necessary for survival, but it may not improve social control, especially in the face of armed militias. Violence in rural Mexico dominated by the drug trade is another context in which cohesion and violence may be related in complex ways. These are important topics for future research.

23. Tocqueville 2000.

24. McQuarrie and Marwell 2009.

25. Small 2009.

26. Robertson 2010. See also Gratz 2010.

27. See especially Marwell 2004, 2007.

28. Ousey and Kubrin 2009; Stowell et al. 2009; Martinez, Stowell, and Lee 2010.

29. The census reports coming out from 2010 show the continuing spread of immigration to the suburbs and countryside across America (Tavernise and Gebeloff 2010).

30. Canada appears broadly similar in this regard. European immigration presents a very different historical context and as a result an apparently different scenario, although the long-term picture is not yet clear.

31. On the importance of simultaneously modeling individual choice and neighborhood change, see Bruch and Mare 2006. See Sharkey 2011 on unselected neighborhood racial change.

32. Deaton 2008.

33. Cartwright 2007. As if to make the point for neighborhood-effects research, consider the unsubtle claim of a recent observer about the Moving to Opportunity experiment: “MTO is the gold standard” (Smolensky 2007, 1016). Educational research has also seen a strong experimental push (Raudenbush 2008), as has the field of criminology, which I explore in more detail elsewhere (Sampson 2010b).

34. Often called “Rubin causality” after the pioneering work of the statistician Donald Rubin, the counterfactual model has become the dominant conceptual framework for casual inference (Holland 1986). For an excellent book-length treatment, see Morgan and Winship 2007. Tools such as propensity score matching, inverse proportional treatment weighting, and instrumental variables are now common in the social sciences as a way to get at causality.

35. See, for example, Kamo et al. 2008; Sikkema et al. 2000; Weisburd et al. 2006.

36. Student mobility and interactional patterns across classrooms thus bear directly on educational experiments (Raudenbush 2008).

37. Wikström and Sampson 2006, 2003. One often hears that experiments are preferable to observational studies because they require fewer assumptions. But experiments just require different assumptions. Depending on the behavior, it may even be that experiments make more heroic assumptions than observational studies, as argued in chapter 11.

38. As the philosopher James Woodward argues: “To the advocate of the causal conception, relations of statistical relevance are beacons guiding our way to crucial explanatory relations” (Woodward 2003, 350, emphasis added).

39. It is important to distinguish between the random clinical trial (RCT) and “field experiments” that may or may not include randomization. Darwin conducted what I would call biological field experiments in his backyard garden (Boulter 2008). The letter-drop studies in this book represent another type of field experiment—there is intervention but not randomization.

40. Lieberson and Lynn 2002, 1.

41. Naive positivism is properly derided in many quarters, as it is perhaps the last holdout that views causality in a deterministic fashion and frowns on the idea of unobserved mechanisms. Most practicing researchers go about their work addressing probabilistic outcomes, thinking about unobserved mechanisms, and recognizing contingencies. I am one of them. I reject naive positivism and “scientism” but accept the aspirations of science, which, broadly defined, I view as a commitment to principles and procedures for the systematic pursuit of knowledge involving the formulation of a problem, the collection of data through observation or experiment, the possibility of replication (science is “public”), and the formulation and testing of hypotheses by a variety of means (science is catholic on method). For an interesting view on causality from the philosophy of science, see Steel 2004 and Reiss 2009. In sociology, see Abbott 1998. My views on causality are probably best described as pragmatist in a critical-realism tradition (Hacking 1999).

42. Abbott 1999. See also chapter 2. The most explicit theoretical statement of human ecology after the original Chicago School was Amos Hawley (1986). But his account fell into some of the same traps as McKenzie’s and perhaps even Park and Burgess’s, with an emphasis on sustenance mechanisms and the physical environment to the neglect of cultural symbols, change, and social mechanisms of reproduction. Yet the generative insights of ecology and the idea of interacting community systems were not lost, eventually migrating into anthropology with important results. I thank Gerry Suttles for this insight.

43. http://www.cdc.gov/mmwr/preview/mmwrhtml/00056796.htm.

44. Smith 2009.

45. Cartwright 2007, 11.

46. By this logic the sort of claims we often hear of quantitative over qualitative data, or vice versa, are unsustainable. More generally, see Becker (1996), who makes the correct call, I think, on the real issue being the quality of the empirical inference (e.g., ethnography does not own the study of “process” or culture, and it can be quite systematic; quantitative methods can be as nonsystematic and nongeneralizable as an ethnography, and so on). In fact I would say that, while quantitative in nature, the methods of this book are inspired by some of the same principles as ethnography, especially the systematic focus on context and social process. Duneier’s (1999) ethnography, for example, is obviously very different than mine but it shares a similar impulse and a systematic method of social inquiry that I believe aligns with the sort of quantitative investigation of the city undertaken in this book. While also very different, Katz’s (2009) ethnographic effort undertakes comparative neighborhood inquiry with a focus on social process. In my research on the life course, I have taken a dual qualitative-quantitative approach to questions of stability and change in the lives of individuals, and so perhaps I see less tension in contrasting forms of data than commonly argued (Laub and Sampson 2003). For a discussion of the role of systematic theory in ethnography, see Wilson and Chaddha 2009. On causal inference and ethnography, see Katz 2001.

47. Lieberson and Horwich 2008. See also their conception of the jury trial as a metaphor for adjudicating evidence in the social sciences.

48. For a description of Snow’s methodology, see Johnson 2006.

CHAPTER SIXTEEN

1. Klinenberg’s (2002) study of two communities in Chicago after the heat wave of 1995 showed how a natural disaster was mediated by social context in explaining the death toll.

2. In reporting on Daley’s decision to not run for reelection, the New York Times asserted that “All cities change over time, but Chicago may be in a class by itself” (Saulny 2010).

3. Pattillo 2007. See also Hyra 2008.

4. Technically, Obama hails from Kenwood near East 50th Street, although many locals consider Hyde Park to include up to 47th Street. What is known as North Kenwood (north of 47th) has traditionally been linked with the poorer community of Oakland.

5. For further details on data sources, see Woodstock-Institute 2010. I also calculated the rate of foreclosures that were filed for the years 2007–9, obtaining nearly similar results in the ranking of high- and low-rate communities. However, Washington Park increased dramatically in foreclosures after the fall of 2008. For 2007–9 overall, Riverdale had the highest rate, and, with Englewood and West Englewood, remains in the top five by either measure.

6. The organization traces its founding to a group of religious and block-club leaders bringing together a coalition of over a hundred neighborhood associations, religious institutions, and civic organizations, aided by Saul Alinsky of community-organizing fame. Over the years the organization grew and mobilized residents for numerous causes. http://www.twochicago.org/.

7. Brazier’s influence in Chicago is widely recognized, especially on the South Side but also nationally. In October of 2010, he passed away at the age of eighty-nine, after which President Obama issued a statement noting that Brazier “promoted spiritual empowerment and economic development through his pastorate of Apostolic Church of God and leadership of numerous community organizations and charitable efforts.” See http://www.whitehouse.gov/the-press-office/2010/10/22/statement-president-passing-bishop-arthur-m-brazier. His legacy lives on, literally. The Apostolic Church of God’s executive board installed Brazier’s son, Byron, as pastor in 2008 after the elder Brazier retired. The younger Brazier appears to be catching up to his father’s level of connections, having served in a high-level post at the Chicago Housing Authority and sitting on numerous boards.

8. See also Sampson and Bartusch 1998.

9. Sweeney 2010. “That’s not the morals and values we live by,” chimed in Marc Robertson, an ordained minister, when referring to the intrusion of violence.

10. Pattillo (1998) reports on an earlier manifestation of spatial vulnerability to crime in a middle-class African American community on Chicago’s South Side, reinforcing again the powerful mechanisms of social reproduction.

11. Sweeney 2010.

12. Spielman et al. 2010.

13. Sweeney, Gorner, and Germuska 2010.

14. Spielman et al. 2010.

15. I completed this chapter section in August of 2010, just after a trip to Chicago and my last visit to the community of Chatham. The July crime statistics were thus the most recent available at the time. The reader can produce his or her own map for any area of the city, although the temporal window for searching crimes is limited to the three months prior to the date of access. See http://gis.chicagopolice.org/CLEARMap/startPage.htm#.

16. http://www.ceasefirechicago.org.

17. Although I debated with myself whether including a link to the video was sensationalistic, I decided to err on the side of including all relevant information, consistent with my liberal use of citations and documentation throughout the book. For a news report on the story that includes the video of the melee and a later protest by local leaders, see http://www.chicagobreakingnews.com/2009/09/derrion-albert-vigil-and-march-postponed.html.

18. The criminological literature at times has alluded to the idea that poor communities tacitly support gangs (see Kornhauser 1978). I saw none of this in evidence on this day, at least, and none in the broader surveys of Chicago residents. Gang violence is widely condemned.

19. For just such a critique from the conservative side, see MacDonald 2010. Of course, the liberal Daniel Patrick Moynihan long ago voiced similar concerns of the plight of black men and family structure in the black community (see the discussion in Massey and Sampson 2009). The issue has been taboo, but with the rise of incarceration it is back on the table.

20. Chicago Fact Book Consortium 1990.

21. A nationally popular song by Jim Croce in the 1970s captured common attitudes:

Well ’ole south side of Chicago

Is the baddest part of town

And if you go down there

You better just beware

Of a man name of Leroy Brown.

So begins a string of racial stereotypes, with Leroy Brown a gun-toting thug who drives a large Eldorado, womanizes, fights, and flashes his diamond rings. Cultural stigma by place is real and, as my findings on the perceptions of disorder show, highly persistent. In 2010, many in Chicagoland would still claim the South Side as the baddest part of town. A larger North/South Side distinction applies to the suburbs as well—the “north shore” is widely deemed to be preferable compared to the south suburbs. I know this firsthand, having lived in southern Cook County for twelve years. Although my kids were raised there and I was (and am) deeply fond of the area, eyebrows were often raised when I declared my residence to those living north of Roosevelt Road (and who often had not ventured south, except perhaps to the University of Chicago).

22. Data released in 2010 from the American Community Survey (ACS) are not up to the present task because they are based on averages from 2005 through 2009 (see chapter 5), meaning that the bulk of the ACS predates the recession. My goal is to examine pre- and postrecession estimates.

23. I examined the disadvantage index of poverty, unemployment, welfare, and female-headed families both with and without percent black. Because racial segregation is so tightly bound up with economic and family status, the results are nearly identical. I graph the full index.

24. Burnside is unusually disadvantaged too, but has the smallest population size of all communities in Chicago (less than 3,500 people in 2000) and so the data are more unreliable.

25. Davis 2010.

26. Neighborhood organizational alliances in New York are said to provide an “old fashioned bulwark in a tide of foreclosures” (Powell 2010). Community-based foreclosure prevention has also been reported to work in Minneapolis (Quercia, Cowan, and Moreno 2005).

27. I examined residential stability (home ownership and tenure length), percent non-Hispanic black, concentrated disadvantage (separating out race), and the density of nonprofits as predictors of foreclosure rates in 2007–9. I also examined 2009 foreclosures, but with fewer filings, the rates were more skewed and not as reliable overall. Controlling for structural characteristics, all of which were measured in 2000, nonprofit density in 2004 was the second-largest predictor of foreclosure filings in 2007–9 (standardized coefficient = –0.35, p < 0.01). The largest predictor was percent black (0.56, p < 0.01) followed by disadvantage and stability, which were also significant (0.24 and 0.22, respectively). Over 80 percent of the variance was explained, confirming that, like the other phenomena examined in this book, the risk of foreclosure varies sharply and systematically by community characteristics. When I weighted the regression model with the square root of the number of HUD estimated mortgages to account for variability in precision of estimates, similar results were obtained. I also examined diversity (immigration), median household income, and changes during 2000–2009 in the index of CHA and housing-voucher units that were unexpected based prior disadvantage. The basic pattern was upheld.

28. Similar results obtain for burglary (0.78), robbery (0.91), and homicide (0.84), all p < 0.01.

29. Specifically, the standardized coefficients are –0.31 and 0.21 (both p < 0.01).

30. Again, my goal is not to present a definitive causal analysis but to ask whether the basic patterns that were uncovered earlier continue. The spatial model is a good test, as the lag of violence measure adjusts for both observed and unobserved predictors of violence in a neighborhood’s risk profile (see Morenoff, Sampson, and Raudenbush 2001). As in chapter 7, however, I believe that ultimately there is a reciprocal feedback relationship between crime and collective efficacy, but this relationship is not estimable with current data. I am pursuing this issue in current analysis.

31. Standardized coefficient = –0.35 (p < 0.01). The CHA/Voucher index of poverty in 2009 was positively and significantly linked to violence in 2010. The number of communities to which a focal area was connected (outdegree) was controlled, but it was not significant. This basic pattern corresponds to the link between the concentration of leadership ties in 2002 and lower homicide rates in 2002–6 that I reported in chapter 14, suggesting a durable organizational component to the social control of violence in a community.

32. The letters were addressed to a person on the “6th Floor” of a building at a street address in Cambridge, Massachusetts (Harvard University was not named). The text was identical to the 2002 study in the business condition letter. The fictitious names of companies and letter-writer signatures (in case the envelope was opened) were randomly varied for added measure. All letters were stamped and addressed with bold type behind water-resistant cellophane. The only unexpected design variation occurred at the end, when I ran out of time and had to drop letters in two communities (Uptown and Irving Park) at night—no night drops occurred in the 2002 study. Return rates were possibly attenuated by street-sweeper rounds in the early morning and the fact that a number of hours passed between the drops and when passersby would appear in the day. Otherwise I strived to drop letters the same way and in a random fashion across contexts—on the sidewalk, near a mailbox, near a parked car door, and near a business, in particular. I repeated this systematic drop routine across blocks within all nine community areas, using a combination of walking and driving, depending on the type of drop (e.g., sidewalk vs. near a parked car). Within days of my return to Cambridge, dozens of letters began to appear. Ninety percent of the eventual returns appeared within a week, and after two weeks the letters trickled off to zero.

33. I therefore adjusted the return rate in all analyses below by calculating the residual from a regression with the nighttime indicator (coded 1 for Irving Park and Uptown, 0 for others) as the sole predictor of return rates. I compare the 2010 adjusted return rate with the adjusted letter-drop return rate for the early 2000s, so that, overall, conditions are not confounded.

34. In this regression the t-ratio for moral/legal cynicism was –2.62 (p < 0.05). I repeated the analysis, controlling for the 2009 foreclosure rate instead of the CHA poverty index, with the same result—moral/legal cynicism from 1995–2002 had a long-term negative association with community-level propensities for other-regarding behavior as revealed in the letter-drop experiment. In similar tests, neither collective efficacy nor the density of nonprofits was directly associated with the 2010 letter-return rate. Although the sample size precludes a definitive analysis, moral/legal cynicism thus stands out among the neighborhood social processes I have considered. For a recent cultural extension of the legal cynicism theory originally set out in Sampson and Bartusch (1998), see Kirk and Papachristos 2011.

CHAPTER SEVENTEEN

1. Kohn 2003.

2. Zorbaugh 1929, 171. Zorbaugh also reported that in the square half-mile vicinity “every year for the past eighteen years there have been from twelve to twenty murders.” For a map of gangland Chicago in 1931, see http://www.encyclopedia.chicagohistory.org/pages/11538.html, where a symbol for Death Corner, centered on West Oak, is shown with the subtitle “50 murders—Count ‘Em” (Grossman, Keating, and Reiff 2004). Some put Death Corner at West Oak and what is now North Cleveland, but which is only about a hundred feet to the east. See also the map in figure 1.2.

3. The reader can see the same site at an undetermined earlier period (I am guessing 2008 or 2009—Google does not date its pictures). The following address was accessed July 28, 2010: http://maps.google.com/maps?rlz=lT4RNWN_enUS305US300&q=oak+and+cambridge,+chicago,+il&um=1&ie=UTF-8&hq=&hnear=W+Oak+St+%26+N+Cambridge+Ave,+Chicago,+IL+60610&gl=us&ei=4aFQTMvfC
oT58AbH5MCVAQ&sa=X&oi=geocode_result&ct=title&resnum=i&ved=oCBMQ8gEwAA. In Street View at the time, the building in my picture can be viewed by looking south. The Cabrini buildings to the northwest are now gone. Across from the abandoned building in figure 17.1 is a still-empty lot, which can be viewed by looking north. A few blocks to the east, one can clearly see the Trump Towers and Loop skyline to the south. To the west are many new housing units, some mixed income.

4. October 13–16, 2010.

5. The faces of the men were drawn in a shape that suggested to me a youthful source. Pictures that I took of these markings are available on request.

6. A modern-day reading of Zorbaugh (1929) repays the effort. He wrote on the increasing “interlocking directorates” (261) among social agencies; community organization as part of revitalizing democracy in the life of city areas (265); segregation and the rise of social distance (242); community perceptions and “mentalities” (244) underlying organizational capacity; the sequential segregation of ethnic and minority groups (127); the lack of dense ties along the Gold Coast (65); why people return to poverty neighborhoods (134–5); spatial distance and the stigma of the South Side to Gold Coast elites—”People in this neighborhood refuse, as a matter of fact, all invitations to tea or dinner on the South Side” (64); and the concentration of leadership ties in the North Side (62). These and other social mechanisms were richly described in ways that bear import, suitably tailored, for today’s city.

7. http://www.dreamtown.com/neighborhoods/cabrini-green.html. As of August 10, 2010, residents still remain in the last Cabrini building “freakishly close to the giant REI store, the glitzy British School and the new Whole Foods cathedral” (Smich 2010).

8. The Urban Villagers is perhaps the most famous sociological account of the unintended consequences of urban renewal in the mid-twentieth century (Gans 1962). In this case an Italian slum was eradicated in the West End of Boston, suggesting a general mechanism at work.

9. See also Sharkey 2010.

10. Skogan and Hartnett 1997. This is not “broken windows” policing as typically portrayed in the media.

11. Western 2006.

12. Durlauf and Nagin 2011. The motive for privileging the police over prisons is found in the longstanding evidence that the certainty of apprehension is much more effective than severity in reducing crime. Prisons have become the poster child for severity of response, as the painting found on the wall of the Cabrini-Green neighborhood described above so graphically revealed.

13. This argument is consistent with the philosophical underpinnings of chapter 9. Although beyond the scope of my purpose here, it is quite possible to conceive of the “good police stop” using the conceptual tools I articulated. Empirically, there is also evidence that when the police work in good faith with local institutions as partners rather than as outside enforcers, crime control efforts are more accepted and more effective. See especially Berrien and Winship 2002.

14. This idea complements an emerging body of scientific evidence (Heck-man 2006; Shonkoff and Phillips 2000) that links children, communities, and schools. See Raudenbush 2009 on school organizational reform and current efforts in Chicago to improve schools and the learning environment within disadvantaged neighborhoods.

15. See Wilson 2010, especially for a description of the multifaceted intervention in the Harlem Children’s Zone. See also http://www.hud.gov/offices/pih/programs/ph/cn/ and http://www2.ed.gov/programs/promiseneighborhoods/index.html. The Millennium Villages project in Africa is an interesting non-U.S. example of empowering villages and improving both health and the physical infrastructure, very much in the spirit of a community-level theory of wellbeing. See http://www.unmillenniumproject.org/mv/index.htm.

16. Both the Moving to Opportunity program and mixed-income policies such as HOPE VI are cases in point. These and similar programs tend to assume a static equilibrium and do not account for the interdependencies among neighborhoods in social mechanisms nor the macrolevel political and social environments that can reinforce segregation.