1. Among the few researchers who have provided data on trends in the numbers of gang cities are Malcolm Klein and G. David Curry. Klein (1995, p. 91) provided figures on the number of gang cities between two periods designated as "before 1961" and "up to 1992." His figures are cited in the "Gang Localities in the United States: A Quarter-Century Report" chapter of the present Report. Curry and colleagues trace changes in numbers of gang cities in several reports, including a section on "Changes in the Gang Problem over Time" (Curry et al., 1992). They compare Curry's own 1992 data on numbers of gang cities with those reported by Walter Miller in 1975 and 1982, Jerome Needle and William Stapleton in 1983, and Curry and Irving Spergel in 1988. Curry also includes several tables in a section entitled "Reported Changes in the Number of Jurisdictions With Gang Problems" (1996). No study available by 1995 reports or analyzes trends in gang counties or includes breakdowns by regions or population categories.

2. See Miller, 1975, 1976, 1982.

3. See Miller, 1982, pp. 4-5. Of the 18 substantive topics listed as "major questions," the present Report directly addresses only two: "Where are youth gangs located?" and "What regions and cities show the highest concentration of gangs?" However, 15 of the 33 tables in the 1982 report present data that deal directly or indirectly with localities.

4. Methodologically oriented readers will note that none of the many locality units counted and tabulated in this Report were obtained by sampling—a procedure widely used in a type of research known as survey research. Sampling is a procedure for ascertaining the size of a set of units without actually counting them. A familiar example is national-level opinion polling, which attempts to determine opinions and preferences of very large numbers of people—often the entire adult population of the United States—by questioning a very small part of that population. Elaborate statistical techniques are used to reach conclusions about the opinions of the larger populations by eliciting the opinions of the people in a small subset. A common type of national opinion poll may base its findings on answers to questions by a subset of a national population group (e.g., all registered voters) as small as 0.00001 or 0.00002 percent of the total population.

Sampling based on a simple random selection of units in the subset to be queried produces the most accurate results when certain statistical criteria are met and when all units in the larger population (the universe) have an equal chance of becoming part of the population subset or sample. It is a useful method of gathering information in situations where the individual identities of the people (or units) included in the larger population but excluded from the sample are not important. For example, the population of registered voters in the United States in 1996 was 127.615 million. When a Los Angeles Times national opinion poll sampled 1,392 registered voters, the identity of 127.614 million, or 99.9999 percent of the registered voters, was unknown.

While sampling may be an appropriate technique in studies where it is not necessary to know the actual identity of all units in a designated population (see, for example, research proposed under "Definition Problems and Accuracy of Information)," this is certainly not the case in the present study, where the objective is to count and identify all known gang-problem localities in the United States. The use of sampling in this situation would produce counts that included many unidentified localities. Knowing the names of reported gang localities is essential to the aims and methods of the present study. The following examples illustrate why sampling is not an appropriate method in the present situation.

The primary audience for this study consists largely of persons who are affiliated with identified States, cities, towns, and counties and whose major concern is the youth gang situation in their own jurisdiction. The listings of all known gang localities in appendixes D and E enable readers to scan a list of names to see whether their localities are included. They can also examine sections of the Report that analyze the seriousness, duration, and prospects of local gang problems and the standings of States and counties ranked according to the numbers of gang localities in each. All these analyses require specific identification of localities.

Another reason that locality identities are needed relates to the analyses of trends and rates. These require knowing the population figures for specific cities, counties, regions, and States—figures that can be obtained only if the names of the localities are known. Locality population figures are essential to the many analyses in the Report that measure changes between the 1970's and 1990's. For example, all of the findings on increases in the numbers of smaller gang cities, and in the numbers of gang localities in the South, require information on their populations in both decades. Assigning localities to specified population categories plays a major role in the analyses in the "Regional Trends in Gang Cities," "Trends in Size of Gang Cities," and "Summary and Explanations" chapters of this Report. An additional reason for identifying all gang localities involves information on the exact geographic location of gang-problem cities and towns. For example, the analysis of predicted growth rates in the "Trend Prospects for Gang Localities" chapter requires information on the location of smaller cities in order to distinguish communities located in or near metropolitan areas (e.g., suburban satellites of large cities) from those located outside of urban areas. Without knowing their names, such an analysis would not have been possible. None of the three objectives just cited could have been achieved if the localities involved were not identified. Thus, sampling would not have been an appropriate technique for the present study.

5. The issue of definition of "youth gang" is treated extensively in other works. Among the more recent discussions are those of Irving Spergel (1995, pp. 16-25) and Malcolm Klein (1995, pp. 20-30). Spergel uses the term "youth gang" to refer to the unit under consideration, while Klein uses the term "street gang." This difference in referring to the major object of concern reflects a longstanding controversy which, at the time of writing, is far from resolution. One of the major sources of contention concerns the role of age in defining gangs; for the proponents of the "street gang" usage, age is not a major consideration; for those using the term "youth gang," age is a major distinguishing characteristic. Miller presents a conception of youth gangs as one type of "law-violating youth group" and a definition derived from interview responses of 300 respondents (1982, pp. 16-22). Curry includes an excellent discussion of the definition issue (1996, pp. 31-34).

6. A comprehensive collection of articles and excerpts relevant to the issues involved in attempting to develop a widely accepted or uniform definition of the term "gang" is contained in a set of documents compiled by NYGC for its Gang-Related Data Collection and Analysis Focus Group (National Youth Gang Center, 1996), whose first meeting was held in Washington, DC, in September 1996. A major objective of this group was to address the issue of what David Curry calls "a universal legal definition of ‘gang.'" This 700-page volume contains 26 separate documents presenting different approaches to the problem of developing a uniform definition. A good example of the flavor and intensity of efforts to achieve a reasonable consensus on the definition issue is contained in Spergel and Bobrowski (1989), an edited transcript of a Chicago conference on developing uniform definitional criteria. Both law enforcement personnel and academicians attended this conference, and the numerous disagreements and confrontations among participants are clearly evident in the published volume. Curry's discussion of "Differences in Definition" (1996) is especially noteworthy. After completing national gang surveys in 1992 and 1994, Curry felt strongly that without a uniform or universally accepted definition of "gang," all of the collected statistical data on gang prevalence, numbers, and other quantitative variables would be suspect because of a virtual certainty that units designated as gangs by different agencies and sources would not be mutually comparable. Curry recommended a renewed effort to achieve a uniform definition. A few years later, however, his hope that such a definition could be developed had waned, and in his 1996 report he wrote, "the struggle for a universal legal definition is commendable, but the legal realities do not make this struggle a promising one."

7. See Miller, 1982, pp. 10-15, for details on characteristics of survey respondents.

8. See Miller, 1982, pp. 20-22. Approximately 1,400 definitional elements were provided by respondents. The six gang criteria cited most frequently were being organized, having identifiable leadership, identifying with a territory, associating continuously, having a specific purpose, and engaging in illegal activities. The three criteria with an agreement level of more than 90 percent were leadership, continuous association, and having a purpose.

9. This figure for the total number of cities in the United States in 1970 and subsequent figures for the total number of counties are taken from table 420, "County, Municipal, and Township Governments, 1972, and Their Population, 1970, by Population Size Groups" (Bureau of the Census, 1974, p. 262). The Bureau of the Census uses several terms and definitions for referring to numbers of cities, towns, and counties. Table 420 designates city units as "municipalities and townships," which are defined as localities with municipal or township governments. The census also uses the term "incorporated places" for places reported to the Bureau as "legally in existence under the laws of their respective States as cities, boroughs, towns, and villages." This usage is found in the table entitled, "Incorporated Places, by Population Size: 1960 to 1990" (Bureau of the Census, 1993, p. 24, table 45). Data derived from both of these usages are employed in this Report. Because of its focus on long-term trends, only those census tabulations that permit direct comparisons between the 1980's and 1990's can be used here. The "municipalities and townships" definition is used for most tabulations, but the "incorporated places" definition is also used. In some cases, differences between tabulations based on the two different types of unit produce statistical discrepancies. For example, on the basis of "municipalities and townships," the total number of U.S. cities and towns in 1992 was 35,935 (Bureau of the Census, 1993, table 420); on the basis of incorporated places, the number was 19,290 (Bureau of the Census, 1993, table 45). The total county population using the municipal unit basis was 224.9 million (Bureau of the Census, 1993, table 473); on the basis of the all-State population, 248.7 million. This discrepancy results from a counting convention used by the census. The census enumerates governed localities for three types of units—counties, municipalities, and townships. In some instances, county and municipal governments are consolidated. In these cases, the census designates the consolidated units as municipal rather than county units, thus reducing the population totals for counties. As a consequence, as noted in footnote 2 to table 5, county population figures based on the municipal government tabulations are 10 percent lower than the total county population reported elsewhere by the census. See the footnotes for tables 4, 5, and 22.

10. Since there are no established procedures for dealing with the city-county issue in the context of counting gang localities, the author submitted these counting rules to several researchers who have conducted gang-locality surveys. All of those contacted confirmed the acceptability of the rules as formulated here. David Curry, who has had more experience in conducting gang-locality surveys than any other researcher, was particularly positive in validating the rules.

11. With respect to this and other calculations in which differences between two or more sets of figures are calculated and reported, the author considered the option of applying tests of statistical significance to the many sets of differing values in this Report. Initial calculations showing time-period differences between comparable units were subjected to nonparametric tests, using the SPSS software package (Release 6.0). All of the initial tests produced significance levels well beyond the 0.01 probability level, and the author decided that such tests would contribute little to the import of the many statistics-based findings in the Report. The large differences between most sets of measures in subsequent tables appear to confirm that using tests to validate statistical significance is unnecessary.

12. One of the few studies that focuses directly on the gang county as a unit was conducted by Claire Johnson and her colleagues for the Institute of Law and Justice (Johnson, 1995). Johnson mailed questionnaires to prosecutors' offices in 368 counties; gang problems were reported for 192. Johnson contacted all offices in the 175 largest counties and sampled smaller counties. There was no attempt to conduct an exhaustive survey of all U.S. counties. Spergel and Curry's 1988 survey (Spergel et al., 1990) includes a list of 22 gang counties as part of their larger survey of gang cities. Using county lists developed in part from the dataset of the present study, John Moore and his associates in the National Youth Gang Center sent survey schedules to 1,300 county sheriff's departments. Of these, 515 reported gang problems in their counties (National Youth Gang Center, 1997, pp. 9-12); 220 of these counties were not included in the dataset used in the present Report.

13. Curry (1996, table 7) reviewed the number of gang cities reported in national surveys between 1975 and 1995. Starting with six gang cities identified by Miller in 1975, Curry traces the number of gang cities, both specified and estimated, reported in subsequent years. Miller's 1982 study was based on a count of 201 identified gang cities, and an estimated total of 286 cities, in 1980. Spergel and Curry reported 68 cities in 1988. Curry reported 110 gang cities in 1992 and 760 in 1994. Klein reported 766 gang cities by the end of 1991 and estimated the existence of 800 or more. Following Curry's tabulation, John P. Moore and his colleagues, using lists developed in part from the dataset of the present study, sent surveys to 2,820 city police departments, 1,492 of which reported gang problems in their cities (National Youth Gang Center, 1997:9-12). Of these cities, 594 were not included in the dataset of the present Report.

14. See note 12.

15. The subregions or divisions of the United States are shown in figure 15.

16. See note 9.

17. The distribution of the cities in the 1970's and early 1990's takes the form of a bell curve, also known as a normal or Gaussian distribution curve. Exploration of a normal probability distribution may be found in any introductory statistics text.

18. The trend noted here, an increase in the numbers and proportions of smaller cities with gang problems, was just getting under way in the late 1970's. In comparing the size of gang cities in States with gang problems, the author was surprised to find small gang cities disproportionately prevalent in California. "If one assumes that . . . developments in California represent the wave of the future for the rest of the country, this may presage a new national development whereby problems with youth gangs will break out of their traditional location in the largest cities and appear with increasing frequency in smaller localities as well" (Miller, 1982, p. 33). Present data confirm this prediction.

19. For a discussion of the reasons for formulating these explanations, see the "Possible Explanations for Growth Trends" section.

20. One of several theories attempting to explain increases in drug trading by Los Angeles gangs in the 1980's was reported in 1996 by Gary Webb (1998), then a reporter for the San Jose Mercury News. Webb's newspaper series and book raise the possibility that the Central Intelligence Agency (CIA) supported Nicaraguan drug dealers who sold crack cocaine to Los Angeles, street gangs in the 1980's to raise money for the CIA-backed Contras, who were fighting the Marxist Sandinista government. The validity of this theory was denied by the CIA and other Federal officials, and both the facts and interpretations of Webb's story were disputed by other writers. Despite this, these allegations were taken seriously enough to produce formal investigations by the CIA, the U.S. Department of Justice, and the Intelligence Committees of both the House and Senate of the U.S. Congress.

21. See Howell, 1996 and 1997.

22. The urban riots of the 1960's and the role of youth gangs in the riots are discussed in some detail in Miller, 1981, pp. 314-315.

23. The role of the Federal Government in granting large sums of money to urban youth gangs, primarily through the Office of Economic Opportunity, which was established by the U.S. Congress to spearhead the Federal War on Poverty, is described in detail in Richard Posten's 1971 book The Gang and the Establishment: A Story of Conflict Arising Out of the Federal and Private Financing of Urban Street Gangs. Figures cited in Posten's book document about $1.6 million in grants to gangs in New York, and Chicago; a single Chicago gang, the Blackstone Rangers, received an O.E.O. grant of $927,341 in April 1968 (Posten, 1971:235).

24. Characteristics of female-based or female-headed households in inner-city communities and their relation to youth gangs are discussed briefly in Miller (1958, pp. 13-14), and in some detail in Miller (1959, pp. 225-229).

25. The earlier term "broken home" was supplanted in later years by terms such as "single parent families," "fatherless families," and "female-headed households."

26. See table 80, "Children Under 18 Years Old, by Presence of Parents, 1970 to 1992" (Bureau of the Census, 1993, p. 64).

27. See table 43, "Household and Family Characteristics of Black Persons: 1990" (Bureau of the Census, 1992, p. 56).

28. A preliminary research design, based on the proposed research described here, was developed and partially tested by NYGC but the research itself has not been completed.

29. William J. Bratton, Commissioner of Police, New York City, quoted in Butterfield, 1996.

30. Jack Levin, Professor of Criminology and Director, Program for the Study of Violence and Social Conflict, Northeastern University, quoted in the Boston Herald, February 20, 1996 (Flynn, 1996).

31. See, for example, Wolfgang, M.E., and Serracuti, F. 1967. The Subculture of Violence. London, England: Tavistock Publications, Ltd., p. 258.

32. DiIulio, Jr., Professor of Political Science, Princeton University, quoted in Newsweek, December 4, 1995 (Morgenthau, 1995).

33. James A. Fox, Dean, College of Criminal Justice, Northeastern University, quoted in Newsweek, December 4, 1995 (Morgenthau, 1995).

34. Miller's (1982) prediction that gang problems would worsen in coming years was derived from a detailed analysis reported in Miller (1975, pp. 55-73). The discussion addressed five major questions: (1) How does the seriousness of current gang problems compare with those of the recent (10-15 years) past? (2) How do respondents in gang-problem cities perceive the future of gang problems in their cities? (3) What are the major factors (e.g., social, economic, demographic) seen by respondents as influencing the future of gang violence? (4) What do population projections for the youth (approximately ages 12-18) category portend for the future of gang and other youth violence? (5) What is the likelihood that gang problems will develop in cities not now experiencing such problems? Since much of the information supporting the problems will worsen prediction was provided by respondents (see tables 24 and 25, Miller 1975, pp. 67-68), a similar analysis is not possible for the present Report. Data with respect to the fifth question, the likelihood of future gang problems in currently gang-free cities, was provided by respondents in the 1995 National Youth Gang Survey (National Youth Gang Center, 1997, p. 15). Responses by 1,379 respondents were as follows: likelihood high, 7 percent; medium, 30 percent; low, 55 percent; none, 8 percent. The opinion of more than two-thirds of these respondents, then, supports a prediction that the increase in gang problems will slow down.

35. As shown in figure 6 and elsewhere, the number of new gang cities listed for the 1970's was 201. Although these cities are designated as new, they could be more accurately designated as new in the 1970's within the limits of the dataset for this study, which began in the 1970's and thus does not distinguish between cities that first reported gang problems during that decade and those that may have reported problems before 1970. Malcolm Klein, in a publication that became available after the present analysis was completed, indicates the presence of 94 gang cities up to 1970 (Klein, 1995, p. 29, figure 4.1). Subtracting this figure from 201 produces an alternative figure of 107 for the number of new gangs cities in the 1970's. Using this figure for the purpose of projection produces the 5- and 10-year projections indicated in the text.

36. See note 9.

37. The following method was used to estimate the proportion of gang cities and towns with populations of 10,000 or less that were not located in areas near or part of larger gang cities. A sample of 71 cities located in 26 States, about one-quarter of the cities with populations of 10,000 and smaller, was selected. DeLorme's "Map-n-Go" program was used to locate these cities. Two categories of cities were distinguished—those within a major metropolitan area and those located within 20 road miles of a larger city with gang problems. Cities were located through a search function. The DeLorme program frames all major metropolitan areas in the United States with black boundary lines. All cities falling within these boundaries were coded as urban area cities. For cities located outside these boundaries, a trip-mileage function made it possible to locate all named localities within 20 road miles of a larger gang-problem city. Between 70 and 75 percent of the searched cities did not fit either of these categories. In addition, the map graphics made it fairly easy to identify villages and small towns in rural and sparsely populated areas.

38. The process by which reduced growth leading to reductions in the number of gang cities could also produce reductions in the overall scope of national gang problems can be illustrated by a hypothetical scenario. If there are 2,000 gang cities in the United States in the year 2000 and a loss of 500 gang cities by 2010, the number of gangs, gang members, and gang crimes in the remaining 1,500 cities would have to expand radically in order to maintain the same numbers as in 2000, let alone increase their numbers. Such radical expansion appears to be unlikely.

39. The 1996 National Youth Gang Survey (National Youth Gang Center, 1999) records a figure of 4,824 gang localities nationwide for the year 1996. This figure is an amalgam of two components: 1,385 identified gang localities and 3,439 unidentified localities. The NYGC total for gang cities includes 898 identified and 2,949 unidentified cities and the total for gang counties includes 487 identified and 490 unidentified counties. The 3,439 unidentified localities, 71 percent of the total, are a product of estimation procedures (see table 1, "Number of Jurisdictions with Gangs in 1996: Reported Versus Estimated," National Youth Gang Center, 1999, p. 8). Comparing these NYGC figures with the three-decade cumulative figures tabulated in the present Report (see the list the "1998 Update of Selected Data" chapter) produces the following breakout: 898 cities identified and 3,847 estimated by NYGC—2,547 cities identified by this Report; 487 counties identified and 977 estimated by NYGC—1,152 counties identified by this Report.

Comparing the NYGC and current Report figures shows that the NYGC figures are 30 percent higher for both identified and unidentified gang cities; 18 percent lower for gang counties; and 23 percent higher for all localities. Comparing identified localities only, the NYGC figure for gang cities is 184 percent lower; for gang counties, 137 percent lower; and for all localities, 167 percent lower. Although these comparisons may be instructive, the two sets of figures are only indirectly comparable. The NYGC survey and this Report employed different methods of data collection, different time periods (1 year versus 3 decades), and different methods of analysis, including a different method of counting gang counties.

40. Postulating a time period during which the number of gang localities was decreasing, the conclusion would be that the more frequent the issuance of new reports, the higher the likelihood of reported decreases in the number of gang localities.