Boston Community Segregation:
How Bad is a Bad?1

Matej Cepl, matej at ceplovi dot cz

1 Introduction

The goal of this article is to test a level of racial/ethnic segregation among the communities of the Boston area. I would like also estimate whether there is an association between racial/ethnic characteristics of different communities and their social, economical and crime situation.

In order to do this analysis, I had to estimate in the first part of the article a distribution of main racial/ethnic communities in the area. The next step in the analysis was to analyze small sample of the local indicators and assessing similarities in their distribution among different communities classified based on the racial/ethnic distribution of such communities.

I have intentionally used in this article very little statistics, because in my opinion it is possible to create quite informative and more reliable picture using less informal means of descriptive statistics (e.$\,$g., ranking of data) and a full-fledged statistical analysis of the Census data require much more sophisticated analysis than I can provide in the limits of this paper.

2 Segregation of the Boston communities

Figure 1: Division of Boston into communities. Source: MassGIS [2002]

Table: Demographic characteristics of Boston Communities. Absolute values (Total, White, Black, and Hispanic) are in thousands, relative (with % in name) in per cents. Source: [MassGIS, 2002]
Total White Black Hispanic White(%) Black(%) Hispanic(%)
Brookline 57.1 44.5 1.3 1.9 78.04 2.24 3.32

Figure 2: Spatial distribution of the black population in the Boston Communities. Source: MassGIS [2002]
Figure 3: Spatial distribution of the Hispanic population in the Boston Communities. Source: MassGIS [2002]

The area analyzed in this article is based on McArdle [2003]'s set of the ``central/high density cities'' limited to the central area of the Boston metropolitan area (see fig:map-Boston). I have added to the set the City of Brookline, in order to avoid selection on dependent variable, because although Brookline is highly affluent neighborhood, it is nevertheless both high-density, centrally located and urban (unless urban is understood as a euphemism for poor). I have also divided the area of the City of Boston itself into separate communities (according to The Boston Foundation [2002a]), because the City of Boston when analyzed as whole hides behind its size all its internal diversity.

I have also limited my analysis mostly to the Boston black community2 in order to make size of my data manageable, and because extraordinary position of the segregation of the black community [Massey and Denton, 1993] makes it legitimate to make its separate analysis. During the analysis I have ignored two communities, which were clear outliers in data: Harbor Islands (where are just some social institutions of the Commonwealth of Massachusetts and no permanent inhabitants) and Fenway/Kenmore (where is even bigger share of students among the population than is common in the Boston area and results are always rather skewed).

The first step in the analysis was to identify parts of the whole area where the representation of the different races/ethnics is higher than average in the whole area. Data from U.$\,$S. Census 2000 MassGIS [2002] were arranged into the table tab:race-rates. Total number of inhabitants total population in a given community are shown in thousands and the column is calculated from the Census variable P007001. Numbers for White, Black, and Latino population are both in the absolute numbers in thousands (variables P007003, P007004, and P007010) and in per cents as a share of white, black (both non-Hispanic), and Hispanic population on given community. Communities with greater share of black and Hispanic population than the average in whole area (15.97$\,$% and 12.41$\,$%) are in bold letters and underlined respectively. For clear presentation of the results I have created also maps showing the spatial distribution of the Black population in the tracts of analyzed area (see fig:black-Boston) and the same map for the Hispanic population (see fig:hisp-Boston).

Both the table and the map show very high level of segregation. There is on the one hand a long curved area in the southeast of the City of Boston between Hyde Park and Jamaica Plain, which contains most of the black population (just Mattapan, Roxbury, and South Dorchester contains 53.24$\,$% of the Black population in whole analyzed area; blacks are 56.69$\,$% of the total population of these communities). On the other hand area of the north and west parts of the City of Boston (Beacon Hill, Central community, and South Boston) seems to be disproportionately more populated by white people. This contrast is most visible on the line between the second most and the second least black neighborhood in the City of Boston: Roxbury and South Boston. Special cases in the city which doesn't seem to fit in the overall city pattern are Charlestown and East Boston in the northern part of the city and West Roxbury in the southern. The former seems to belong more to the neighborhood of Cambridge and Somerville, and the latter is according to Boston Redevelopment Authority [2001] de facto part of suburbs.

Comparison between the city of Boston and other communities in our area shows that whole city of Boston has much more minorities than rest of the analyzed area. All non-Bostonian communities in our sample have sub-average representation of the Black population and even the highest among these, Cambridge, has just slightly more than half of the average Bostonian share of the Black population (moreover, the map shows that only remarkable black community lives in the projects of the Northern Cambridge).

The situation is slightly different with the Hispanic community. Although there is a similar pattern of local concentration of the Hispanic population in the area of the East Boston, and Chelsea, the level of segregation of Hispanic population is less severe than in the case of the Black population. Eight communities in the area have supra-average share of Latinos comparing to only six Black ones and only 35.71$\,$% of all Latino lives in the three communities with the highest proportion of Hispanic population (East Boston, Chelsea, and Roxbury) and they constitute only 35.26$\,$% of the population in these communities. Moreover, although I cannot prove it from my data (which are all from the same time period), according to Massey and Denton [1993, p. 77] the segregation of the Black communities is usually more perpetual than segregation of other groups and newly immigrated populations.

Standard measurement of the interracial/ethnic segregation is according to Massey and Denton [1993, p. 46] the index of residential dissimilarity (percentage of the minority population which would have to move in order to make racial/ethnic composition of the city even, i.$\,$e., that every unit of the city will have the same proportion of the minorities as whole city). Indices of residential dissimilarity for the analyzed area is 51.72$\,$% for black population and 36.85$\,$% for Hispanic population (just for the City of Boston the numbers are 48.66$\,$% and 34.50$\,$%). My numbers cannot be unfortunately used to compare the residential segregation between the communities with the data of Massey and Denton [1993] (because their indices are based on substantially less homogeneous wider base, they correctly show that the true segregation in the Boston MPSA is substantially higher). However, these my calculations further validate the conclusion, that the segregation of the Hispanic population is less severe than the Black one in absolute numbers.

My data does not cover any time period, so it is not possible for me to analyze any dynamical changes in the segregation. The similar data were analyzed by McArdle [2003] and she got very similar dissimilarity indices of the segregation in the City of Boston (see Figure 10 p. 23), but moreover she analyzed also dynamics of desegregation and found a pattern of change common for both Bostonian suburbs and the city, where a segregation of blacks although higher in absolute numbers is declining more rapidly than the segregation of both Latino and Asian population.

3 Socio-economical characteristics

Table: Economic and Social characteristics of Boston Communities. (*Data for Malden are from year 2000.) Source: [MassGIS, 2002] and the last two columns [The Boston Foundation, 2002b; Bureau of Justice Statistics, DOJ, 2002]
Rate of unemployment
Per capita income
Poverty rate
Viol. rate 2001
Prop. rate 2001
Brookline 2.4424 45,23925 9.2821 0.3726 263.223 1,810.823

After delimitation of the area of the black community itself and assessing the level of segregation in the area, we can continue in analyzing some of its social and economical characteristics.

Before proceeding with the analysis of the data, one note is necessary. I am very much aware of the fact that ``Association does not mean causation'' and so whatever are my conclusions about relations between racial/ethnic composition of communities and crime indicators, these associations should not be under any circumstances used as characterizing particular race or ethnic group as such.

I have used for the analysis data from the U.$\,$S. Census 2000 [MassGIS, 2002]: level of unemployment (P043014 and P043007 for the male and female unemployment together with P043001, total civilian workforce), per capita income in 1999 (P082001), poverty data (P087002 for number of people living under the level of poverty and P087001 for total population for which the poverty is calculated), and drop-out numbers (P037003 and P037020 for share of males and females in age 25 without completed schooling, and P037001, which is total number of population in the age 25). I have also added indicators of the violent crimes and the property crimes (per 100,000 persons in population) from Bureau of Justice Statistics, DOJ [2002] and The Boston Foundation [2002b]. The Census values were calculated into rates for the individual communities and then all data are presented as tab:race-economics.

In order to make sense of the data in the table, I have created rankings of all communities according to the respective columns (indicated by the superscripted number) and then marked with boldface font all values, which were in the upper half of the ranking (summary data on the whole City of Boston were not ranked and distinguished by italic font). Because I want to analyze relation between these data and the racial characteristics of the individual communities, I have marked the ones which were identified in the previous section as predominantly Black by boldfacing their name in the first column and predominantly Hispanic by underlining it. For some variables Harbor Islands and Fenway are clear outliers and in such a case the value is stroke out in the table. Moreover, crime indicators are not available for Harbor Islands.

Looking at the data from a broader perspective, a very clear pattern emerges: all communities marked as predominantly Black or Hispanic tend to fall to the lower half or even to the bottom of all rankings. Specifically the group of communities positioned in the table between (and including) East Boston and Mattapan (which are also geographically neighboring one another) create the bottom of all rankings. And among these Roxbury (second ``most black'' community) stands out as the very worst case in per capita income, poverty rate, and the second from the bottom in violent crime. Another evidence of the association between racially/ethnically segregated communities and poor living conditions is Chelsea which is overwhelmingly Hispanic (48.35$\,$%) and ranking at the bottom with the above-mentioned group and the worst in drop-out rates and violent crime. Surprisingly it stands slightly better in economic characteristics (especially the rate of unemployment).

When we take a more detailed look we can find many exceptions from this pattern. One of them is caused by the huge student population in Boston (approximately 250,000) which skews especially the rates of unemployment (most notably Fenway, Cambridge, and Central), and of course drop-out rates (here it shows most with South Boston, the seat of the University of Massachusetts in Boston, which would be expected according to other indicators to stand much worse than it does). However, we have to be aware of the crudeness of the methodology and limitations of our data. The example of this is very poor position of Chelsea exactly in the drop-out rates. We cannot be sure whether this value of the drop-out variable can be explained only by the suggested influence of the low socioeconomical status of the Hispanic population, or other factors influences the drop-out rate as well. One such possible explanation is that education of Hispanic population born (and educated) outside of the United States is incompatible with the standard American education and thus unacknowledged in the Census. Another comment for the drop-out value is necessary--its range is so narrow, that the rough methodology of analysis could be affected discover possible errors.

There is couple of another exceptions to the general rule, that segregated communities suffer worse socioeconomic characteristics. When considering crime-related indicators, Fenway/Kenmore, Back Bay/Beacon Hill, and the Central community being apparently one of the most affluent by other measures stand out as three worst in terms of the property crime and in the lower half in terms of the violent crime (and the Central community the third worst just following Chelsea and Roxbury). The probable solution to this issue is twofold. First of all, crime does not follow the place of living (as other Census statistical variables do). Second, the magnitude of the difference between Back Bay and Central community, can be explained only by the fact that these communities contain most of the financial district, where certainly most of the property-related white-collar crimes happen.

Another important exception from the rule is Hyde Park. Although it firmly stands in the minority part of Boston by its racial/ethnic characteristics (and geographical location) with 39.07$\,$% and 13.50$\,$% share of the Black and Hispanic population respectively and according to maps consistent minority residency across the area, all its socio-economic indicators except of one are in the upper half of the rank. Further research would be probably necessary to prove it firmly, but it seems that in this area we can find another level of segregation of the already segregated community--this time the area of the Black middle class separated from the rest of the Black community.

Another exceptional community is the South End, which has extraordinary high per capita income (fifth best), but ranks in the bottom of all other variables. I do not have a good explanation of this phenomenon available, aside from the problems with my too crude methodology or missing some data. One variable which should be probably investigated further is the share of foreign-born members of a community, which may explain this and some other irregularities.

Last group of remarkable communities is group in the southwest part of the City of Boston (Jamaica Plain, West Roxbury, and Roslindale), which in all its characteristics puts them outside of the Boston urban area more towards suburbs (Jamaica Plain has a divided nature, with a part falling into the Black community and other one to the semisuburban part of Boston).

4 Conclusion

There are two possible conclusions from my research. One is the direction where the further research should follow. Of course, it may be possible to add another dependent variables to the analysis in order to better explain some surprising results (e.$\,$g., immigrant status). Unfortunately, adding every new variable to the model, would make descriptive method of the analysis even more inadequate and more complicated statistical tools would have to be used. However, it is not the most important objection I have against this step (of course, there is no inherent problem in calculating more complicated models). By creating the system more complicated in order to get better $R^{2}$ I am afraid to loose any possibility of getting understanding and insight into mechanism which plays between community segregation and socioeconomic characteristics. Given the highly intercorrelated character of the most community indicators, it is probably not possible to create unidirectional mathematical model predicting socioeconomic indicators from any number of dependent variables (like race of the community or the amount of governmental support to underdeveloped communities). Therefore, I think that I would rather keep the number of my variables very modest (maybe just add one or two), but I would try to observe effects of the racial segregation on the bigger number of unrelated communities in other metropolitan areas. Another possible direction of the further research may be further rethinking of what exactly means segregation and what is the desired state of the community (is it just zero dissimilarity index?). I have just assumed in this article, that segregation is bad and less it is there better. However, I am afraid that in the further research, so simple theoretical assumption would not be sufficient.

Other conclusion of this articles is more practical. First of all, it is necessary to repeat, that the level of community segregation is still very high even after many years since it was officially outlawed. However, given my insufficient understanding of the desired direction of the inter-communal relations, I cannot suggest any exact policy proposal to follow. I just have to agree with conclusions of Massey and Denton [1993], that the discussion about the means for dealing with communal segregation is clearly insufficient and that most of the current attempts are both ineffective and hugely wasteful. I would also agree, that probably currently the best what we could do is to support small-scale efforts, voluntary organizations, indirectly support minority population through subsidies for rent and building outside of the minority communities, and certainly a vigorous fight against any hate-crimes or expressions of hate against any minority.


Boston Redevelopment Authority.
Population by census tract including race and ethnicity in Boston and Boston's neighborhoods for the year 2000 (PL 94-171 initial release).
In Boston Population-2000. 2001.
Bureau of Justice Statistics, DOJ.
Data online, 2002.
Douglas S. Massey and Nancy A. Denton.
American Apartheid: Segregation and the Making of the Underclass.
Harvard University Press, Cambridge, MA, 1993.
ISBN 0-674-01820-6.
Massgis - datalayers from the 2000 U.S. Census datalayer description, 2002.
Nancy McArdle.
Race, place, and opportunity (Boston).
Study, The Civil Rights Project, Harvard University, April 2003.
two parts.
The Boston Foundation.
Boston Indicators Project.
The Boston Foundation, 2002a.
The Boston Foundation.
Trends in reported Part One Crime per 1000 population, by Boston neighborhood.
In Boston Indicators Project The Boston Foundation [2002a].
Wendy Murray Zoba.
Separated and equal.
Christianity Today, pages 14-24, 2 1996.


Version .
I would like to thank for the advice I got from professors Leonard Buckle, Gordana Rabrenovic, and Glenn Pierce of the Northeastern University. Their encouragement made me to change the mess this article used to be into something hopefully at least partially intelligible.
On the technical side, I owe a huge debt to the several communities of users and developers of open source programs with which I have created this paper: Linux (which is the foundation of it all), LyX (a tool with which this is written), GRASS (full-size GIS package), and PostgreSQL (database where data resides).
I am humbled by all this support and I wish I made better result with all this advice and tools freely given.
... community2
I use mostly term ``Black'' instead of ``African-American'', because the latter confuses my analysis by adding artificial distinction between descendents of the former American slaves and latter immigrants from Africa and Caribbean area (see Rev. Eugene Rivers interview in Zoba [1996] for the similar reasons).

Matej Cepl 2003-12-23