- Open Access
The complexities of measuring access to parks and physical activity sites in New York City: a quantitative and qualitative approach
© Maroko et al; licensee BioMed Central Ltd. 2009
- Received: 12 March 2009
- Accepted: 22 June 2009
- Published: 22 June 2009
Proximity to parks and physical activity sites has been linked to an increase in active behaviors, and positive impacts on health outcomes such as lower rates of cardiovascular disease, diabetes, and obesity. Since populations with a low socio-economic status as well as racial and ethnic minorities tend to experience worse health outcomes in the USA, access to parks and physical activity sites may be an environmental justice issue. Geographic Information systems were used to conduct quantitative and qualitative analyses of park accessibility in New York City, which included kernel density estimation, ordinary least squares (global) regression, geographically weighted (local) regression, and longitudinal case studies, consisting of field work and archival research. Accessibility was measured by both density of park acreage and density of physical activity sites. Independent variables included percent non-Hispanic black, percent Hispanic, percent below poverty, percent of adults without high school diploma, percent with limited English-speaking ability, and population density.
The ordinary least squares linear regression found weak relationships in both the park acreage density and the physical activity site density models (Ra2 = .11 and .23, respectively; AIC = 7162 and 3529, respectively). Geographically weighted regression, however, suggested spatial non-stationarity in both models, indicating disparities in accessibility that vary over space with respect to magnitude and directionality of the relationships (AIC = 2014 and -1241, respectively). The qualitative analysis supported the findings of the local regression, confirming that although there is a geographically inequitable distribution of park space and physical activity sites, it is not globally predicted by race, ethnicity, or socio-economic status.
The combination of quantitative and qualitative analyses demonstrated the complexity of the issues around racial and ethnic disparities in park access. They revealed trends that may not have been otherwise detectable, such as the spatially inconsistent relationship between physical activity site density and socio-demographics. In order to establish a more stable global model, a number of additional factors, variables, and methods might be used to quantify park accessibility, such as network analysis of proximity, perception of accessibility and usability, and additional park quality characteristics. Accurate measurement of park accessibility can therefore be important in showing the links between opportunities for active behavior and beneficial health outcomes.
- Ordinary Little Square
- Geographic Information System
- Kernel Density Estimation
- Geographically Weighted Regression
- Environmental Justice
Environmental justice is the fair and equitable distribution of both the environmental "bads," such as hazardous waste sites, and the environmental "goods," such as parks, open space, and recreational opportunities. For more than a decade, Geographic Information Systems (GIS) have been used to examine the spatial realities of environmental justice [1–11]. GIS methods have been applied in environmental justice research primarily in the analysis of the spatial relationships between sources of pollution burdens and the characteristics of potentially affected populations. Environmental justice research has therefore focused on analyzing the disproportionate exposure of pollution on communities comprised of vulnerable groups, such as racial/ethnic minorities and socio-economically disadvantaged groups, and the concomitant effects of this pattern on health and environmental disparities [12–14]. GIS has been less often used to analyze the relationship between socio-demographic and environmental "goods," such as health-promoting land uses and positive aspects of the built environment.
Previous studies have documented that proximity to parks and open spaces has a positive influence on engaging in active behaviors, like walking and running for exercise [15–19]. Other studies have analyzed how the availability of outdoor space impacts on specific health outcomes, like community-level rates of mortality, cardiovascular disease, diabetes, and obesity [20–26]. The underlying hypothesis is that since individual-level risk factors for these highly prevalent diseases do not fully explain disparities in their distribution across population groups, or even disparities across population groups in health behaviors that are related to these diseases, modifiable environmental factors may help us to develop fuller models explaining health disparities in these health outcomes and related health behaviors. This research is of interest to public health and policy analysts who are developing interventions and policies that can mitigate health disparities that persist across socio-economic groups in the USA.
If environmental factors help us to understand the distribution of health outcomes in the population, then one might expect that active outdoor space would be less available to populations with overall worse health outcomes. Since low SES populations and racial/ethnic minorities experience worse health outcomes in the USA [27, 28], access to parks and physical activity sites becomes an environmental justice issue. However, research findings have been contradictory, which suggests a complex relationship among socio-demographics, outdoor space, and individual-level health factors.
Summary of selected park accessibility research.
Study Area(s)/Unit of Analysis/Independent Variables
Abercrombie et al, 2008
Study Area: Metro Baltimore/DC area (MD)
Unit: census block groups
Independent Variables: % minority, median income, pop size, geographic size, and % pop < 18.
Number of private rec. facilities and public parks per block group; size of rec. space. Number of parks and facilities were recoded into categories based on # per block group and the park size was divided into four categories based on Mertes and Hall's classification system.
Neighborhoods selected by variation in walkability and median income. Socio-demographic variables in tertiles (low, medium, and high). Two way analysis of covariance: # private facilities, # parks, & largest park size across block groups.
No signif. effect of income or % minority on # private rec. Mixed-race neighborhoods had highest number of parks, regardless of income. Low- and middle-income pop. in white block groups and high-income groups in minority block groups had lowest park access.
Estabrooks et al, 2003
Study Area: small American Midwestern city (not specified)
Unit: census tracts
Independent Variables: % unemployed, per capita income; % pop. below poverty threshold, education (less than h.s. diploma). Racial/ethnic characteristics
Availability of PA resources and accessibility as pay-for-use and free-for-use. Raw counts of numbers of PA facilities per census tract.
Multivariate analyses of variance of PA resource availability and accessibility by neighborhood SES; Univariate analyses of variance to determine whether income differed on the number of pay-per-use and free-for-use facilities.
Low- and medium-SES neighborhoods have signif. fewer PA resources than high-SES neighborhoods. Low- and medium-SES neighborhoods have signif. fewer free-for-use resources than high-SES neighborhoods
Moore et al, 2008
Study Areas: Forsyth County, NC; Manhattan & Bronx, NY; Baltimore City & County, MD
Unit: census tracts, blocks, and 100-meter grid cells (kernel density)
Independent Variables: total pop, racial/ethnic pop, land area, median household income.
Presence of resources, as well as densities & types of resources. Public-use parks, commercial and public rec. The total number of resources obtained by summing the resources at each location, weighted by the count when appropriate.
Kernel density of recreational resources, weighted by # of resources and types; binomial regression for probability of having access as function of SES and demographic factors.
Minority & low income areas signif. less likely to have fee-for-use rec. Densities of public rec. within parks were signif. higher in minority and low-income tracts, even after adjustment for pop.
Study Area: Bryan, TX
Unit: census tracts
Independent Variables: Pop density, % non-White; % black; % Hispanic; % < 18; % > 64; % renter occupied housing units; mean housing value; mean rent
Equity and accessibility to parks: ease with which a site can be reached and fairness of distribution of parks.
Buffering Euclidean & street network distance for accessibility; comparison of pop. factors of areas w/good access to pop. factors in areas w/o good access
Large areas of the city are not within 1/2 mile of a park access point, by either the straight-line or network distance. < 40% of pop has good access. All pops seem equally well-served by the parks, and the parks are well-distributed amongst less advantaged groups
Study Areas: Pueblo, CO, and Macon, GA
Unit: census blocks
Independent Variables: % non-white (Macon); % Hispanic (Pueblo); % < 18 years; vacant units; owner occupied units; Median housing value; % housing units w/> 1 person per room; % households w/no spouse present
The spatial clustering of park access scores with the spatial clustering of SES variables. Also used a measure of accessibility at the census block level based on amounts of park acreage within certain distances of residential areas.
Access measure consists of the total amount of park acreage located within a specified travel distance between each block and each park, using street network distance between centroids of blocks and centroids of parks.
Spatial autocorrelation for both cities is significant for park access measures. Park access in Pueblo favors higher-income areas. In Macon, access to parks tends to favor lower-income areas.
Talen and Anselin, 1998
Study Area: Tulsa, OK
Unit: census tracts
Independent Variables: % pop < 18; % non-white; median housing value
Spatial distribution of playgrounds using the shortest path distances over street network from census tract centroids.
Compares the results of "container method" w/the geographic access measures obtained by gravity model (travel cost measure).
The playgrounds are not distributed evenly throughout the city, but are also not predicted by any specific socio-demographic variables.
Timpiero et al, 2007
Study Area: Melbourne, AU
Unit: postal districts
Independent Variables: Index of relative SES disadvantage (income, education, occupation, family composition, dwelling structure).
Density and area of various categories of open space in relation to SES within each postal district.
container approach was used correlating numerous variables (# of OS facilities, OS area, OS density, etc) to SES index
Greater # of o.s in lowest SES districts; once normalized by pop, differences not signif.
Wolch et al, 2005
Study Area: Los Angeles, CA
Unit: census tracts
Independent Variables: Total pop; racial/ethnic pop; pop < 18; median household income; % persons in poverty.
Park access = park acres/1,000 pop (total pop and < 18 pop); % of tract pop (total and < 18) within 1/4 mile of a park boundary; Park acres/1,000 pop (total and < 18) living within the 1/4 mile buffer.
1/4 mile buffers around parks creating accessible park acreage per census tract. Estimates of total area within a 1/4 mile of park and total accessible population per tract were calculated.
Low-income and concentrated poverty areas have relatively low levels of park resources and accessibility. African American,, Latino, and Asian American pops have low rates of park access compared to white-dominated areas.
One of the most common methods used in examining access to park space is called the "container approach." This approach measures access by determining whether or not there is a park or recreational facility within a particular geographic unit of aggregation (e.g., zip code, census tract, or neighborhood), rather than using or developing an actual proximity measure such as Estabrooks, et al, 2003 . In this container approach, the number of parks per areal unit is then summed and associations between this count and various population characteristics, such as SES, can be estimated for the chosen unit of aggregation.
A number of previous studies have used this "container" method for evaluating whether or not a person has good access to parks [29–31], and have found contradictory results when correlated with neighborhood SES (lack of access to parks positively correlated with low SES versus negatively correlated with low SES). These inconsistent results may be due to the container method itself. In addition to the boundary problem discussed above, this method is problematic in that it often does not take into account the underlying population structure and density of the areal unit, or the relative size of these areal units. Thus it is not a fair comparison since areal units with larger populations require more parks than an equivalent areal unit with fewer people in order to be equitable.
Additionally, although some of these studies use actual park acres per geographic unit in their calculations, others simply count the number of parks and facilities per geographic unit. However, creating a simple count of parks and basing equity analysis on that count does not consider the actual amount of park space available to residents, since one park may be substantially smaller than another and therefore should not receive an equal weight in the calculations.
Other studies have used proximity analysis based on "walkability" distances, which is a more refined measure of access, by setting certain distances to parks as a proxy for access, such as 1/4 mile (~400 m) or 1/2 mile (~800 m) as a standard walking distance [32, 33]. However, access by proximity or distance often does not take into account the actual street network, merely Euclidean distance.
There have been a few studies using the street network to calculate distance to parks, for instance, Nicholls (2001) and Talen and Anselin (1998) each of whom compared straight-line distance with street network distance [34, 35]. The Nicholls study found that approximately 80% of the area studied (Bryan, Texas) were not within 1/2 mile (~800 m) of any park (as measured by network distance), but that the less affluent neighborhoods tended to be better served by parks than the more affluent areas. In the Talen and Anselin study, the distribution of playgrounds in Tulsa, OK, as measured by various methods including the network analysis, could be considered "unpatterned inequality." The playgrounds were not distributed evenly throughout the city, but were also not predicted by any specific socio-demographic variables.
Kernel density estimation, or kernel smoothing, is another method for measuring accessibility. "Kernel density estimation involves placing a symmetrical surface over each point, evaluating the distance from the point to a reference location based on a mathematical function, and summing the value of all the surfaces for that reference location. This procedure is repeated for all reference locations."  Kernel density estimation creates a statistical surface so that, for instance, there is an accessibility value as measured by park density, mapped at every point in the study area.
Kernel density estimation is typically considered a more refined spatial statistical model than the container approach. It can give an estimation of accessibility for every point in the study area, not just a binary answer of "within walking distance" or "not within walking distance," as in both the fixed-distance proximity and the network analyses. There have been very few studies of park access using the kernel density method, although this method has been used extensively in other types of analyses [37, 38]. Moore, et al. (2008) used the kernel density estimation method to compare park access in three USA locations, and found that although pay-for-use recreational venues were more likely to be located in white and more affluent neighborhoods, public parks tended to be more equitably distributed, and densities of recreational facilities within parks were significantly higher in minority and low-income census tracts than in white and higher-income tracts, even after adjustment for population .
This analysis uses the kernel density estimation approach to test whether access to park space is associated with neighborhood race/ethnic composition and SES in New York City. Densities of both park acreage and physical activity sites are mapped and correlated with SES measures. The two main categories of data used in this analysis were park information and socio-demographic information.
Data – parks
The park features data (elements within the parks) were created through a collaboration between the New York City Department of Parks and Recreation and Lehman College of the City University of New York. Researchers traveled to all of the New York City parks carrying portable GPS units and recorded the locations of many of the parks' features, including items such as drinking fountains, comfort stations (rest rooms), flag poles, stairways, historical markers, statues, beaches, courts, ball fields, and other recreational areas (see figure 2). This point data (latitude and longitude) were rectified with aerial photos and further processed into a more accurate and useable dataset.
To prepare the data layer for physical activity sites, each discrete non-linear park feature that was identified as activity promoting (i.e. something that encourages caloric expenditure) was extracted from the main parks database, converted to a point (if necessary) and given a value of '1' (1 = physical activity site). The features that were coded are: basketball courts, handball courts, tennis courts, volleyball courts, multipurpose courts, soccer fields, baseball fields, football fields, swimming pools, hockey rinks, golf courses, and running tracks. Kernel density estimation was again performed, this time in order to estimate the density for physical activity sites in New York City. A 1.6 km radius was used as the bandwidth and a 50 meter resolution raster surface was created (see figure 3b).
These two density surfaces were used as proxies for access to park space (acres) and active recreation (physical activity sites), following the assumption that where there are higher densities of resources, access is greater.
Data – Socio-demographics
Joining parks data with demographic data
To link the raster park data with the vector socio-demographic data, zonal statistics were used within the spatial analyst extension of ArcGIS. This process aggregates and statistically summarizes the values of the raster cells whose centroids fall within the corresponding block group. As a result, each census block group in New York City is given an average value for the park acre density and physical activity site density of the grid cells that fall within its boundaries. This aggregates the kernel density-derived statistical surfaces (acres and physical activity sites) to the same geographic unit as the socio-demographic data.
Two discrete statistical methods were used on the data: ordinary least squares linear regression (OLS) and geographically weighted regression (GWR). The datasets used in the analyses were identical. Census block groups with fewer than 256 residents were excluded (lowest 5%) in order to stabilize the model. Block groups that had missing data from any of the variables were also excluded (~ 0.1%). After the data were cleaned and prepared, 5,439 block groups out of the original 5,732 remained (94.9%). Log10 transformed park acreage density (ACRE) and log10 transformed physical activity site density (PAS) were used as the dependent variables. The independent variables included: percent non-Hispanic black, percent Hispanic, percent of adults with no high school diploma, percent below poverty, percent with limited English language ability, and population density.
OLS Regression t-values.
% non-Hispanic black
% no high school
% below poverty
% limited English language
PAS GWR model parameter summaries.
% non-Hispanic black*
% below poverty*
% with no high school diploma
% limited English language*
ACRE GWR model parameter summaries.
% non-Hispanic black*
% below poverty*
% with no high school diploma*
% limited English language*
The GWR analysis identified the relationships between park access measures and socio-demographic variables as behaving inconsistently across New York City. These idiosyncrasies could benefit from a qualitative evaluation of the relationship between SES characteristics with physical activity sites and park density.
Justifications for qualitative analysis
The GWR analysis revealed a tendency toward what has been termed "unpatterned inequality," meaning that while the parks and physical activity sites are not evenly distributed across the city in a geographical sense, neither are they predicted globally by race/ethnicity, income or the other variables typically investigated in environmental justice analyses. There are no discernable consistent associations between park access and socio-demographic indicators . This is, of course, not to say that everyone has equal access to parks and physical activity sites, or that all neighborhoods have good access to these resources. Certainly, a quick glance of a map of New York City's parks indicates that not all parts of the city are equally well-served by parks.
Therefore, we thought it would be beneficial to investigate case study areas on a more detailed basis, which might illuminate the spatial incongruities that exist. More explanatory power may be realized through a qualitative analysis, which includes historical background of the parks and surrounding neighborhoods, a description of the past and current socio-demographics, and an overview of the physical aspects of the study areas.
New York City has a complex relationship between its physical infrastructure and its population distribution. Many parks are quite old and were established in neighborhoods having very different socio-demographic characteristics than those of the same neighborhoods today. Parks are a special category of fixed infrastructure for that reason – most occupy large parcels of land and once they are established, it is unlikely that they will be eliminated or moved. The populations surrounding them, however, are quite changeable. This is why it is difficult to ascribe environmental justice implications to the locations of parks. It is still instructive to try to determine which populations, if any, are currently underserved by parks and recreational spaces in New York City, despite the original purpose of the parks, and who those parks were intended to serve.
Selection of study areas/delineation of study area boundaries
Highland Park and Marine Park Comparison
Park Construction (Year)
Park Size (Acres)
Physical Activity (PA) Sites
Additional PA Sites
bocce courts, cricket fields, hiking trails, skate park, kickball courts, and kayak/canoe launch sites
Capital Projects (money spent since 1995)
Over $6 million
Over $34 million
Sample of Capital Projects
New baseball field and tennis court lighting, added safety measures, synthetic turf soccer/football field (under construction), etc
Landscaping, construction of separate environmental and community centers, golf course irrigation, comfort stations, etc.
Quality of the Park
Good quality overall (see figure 7e): well maintained older play equipment with safety surfaces (see figure 7f), broken chess table (with caution tape), playground safety signs***
Good quality overall (see figure 8c): some graffiti, play equipment with safety surfaces (see figure 8d), "Clean up after yourself" signs, safety signs at every playground ***
Public School children maintain various gardens throughout the park (figure 7d)
An 800 m buffer, generally accepted as the upper end of "walking distance" , was drawn around the boundaries of each park to create the study areas. Census tracts that intersect this boundary were chosen for demographic analysis to represent the approximate catchment area of the park. The intersection was first performed by the GISc software, and the tracts which intersected but had a very small proportion of their area within the buffers were manually removed. For Highland Park, this area includes parts of the following neighborhoods: Highland Park and Cypress Hills, Brooklyn; and Glendale, Queens. For Marine Park, this area includes parts of the following neighborhoods: Gerritsen Beach, Sheepshead Bay, Marine Park, Flatlands, and Mill Island (all in Brooklyn).
In 1891 Brooklyn purchased the land surrounding the Ridgewood Reservoir (built in 1856) to be used as a park. In 1905, the park was extended south, and by 1908 the park was extended west and set the boundaries that remain today. By 1908 the park already included football fields, baseball fields, tennis courts, several gardens, footpaths, and park structures. Situated among a chain of seventeen mid-19th century cemeteries that straddle the Brooklyn-Queens border, Highland Park greenery blends in well with what is known in New York City as the "Cemetery Belt." 
Construction began on Marine Park in 1936 nearby undeveloped marshland around Gerritsen Creek. By 1937, the park included 1822 acres, in large part due to fill deposited in the marshes in the 1930s. With anticipated development, speculators purchased real estate along the waterfront. The vision of a new park inspired home building in the area which included a golf course built in 1963. In 1974, 1024 acres were transferred to the Gateway National Recreation Area. 
Description of physical aspects of study areas
Both of the neighborhoods surrounding both Highland Park and Marine Park are heterogeneous in terms of levels of maintenance, upkeep, and cleanliness and there is a dramatic range of housing types, from mansions to low-income public housing complexes, found within the 800 m buffer of the parks, albeit not necessarily adjacent to one another (see figures 7h, 8f and 8g). There are also various types of business and commercial strips in both study areas (see figure 7g and 8e).
For both parks, it is difficult to qualitatively assess the association between the apparent socio-economic status of the neighborhood and accessibility to the park. For instance, the north side of Highland Park in Queens has much less access to the park than the comparatively less affluent neighborhood to the south of the park in Brooklyn. The Queens neighborhood is isolated from the park by obstacles such as cemeteries and a major highway (see figure 7b and 7c), and is by far the furthest neighborhood from the park's physical activity sites. One anomaly is that the mansions and larger homes located to the west of Highland Park appear to have excellent park access. It is interesting to note that when observing the Brooklyn side of the park alone, housing conditions do seem to deteriorate the further one travels away from the park.
For the Marine Park study area, park access appears more evenly distributed, regardless of SES. While the higher value housing around Marine Park is located near the main physical activity sites (see figure 8h), there is also high access to other physical activity sites from low-income housing complexes. There are some neighborhoods within the 800 m buffer of Marine Park that also have physical obstacles to access. The park itself contains natural areas of salt marsh and streams which are not easily traversable. There are also two neighborhoods in close proximity but separated from the park by water bodies (see figure 8b). One of these neighborhoods is a relatively higher income area and the other is lower-income.
Socio-demographic characteristics of study areas
Since the construction of Highland Park and Marine Park, in 1891 and 1936, respectively, the demographics of the neighborhoods surrounding these two areas have changed considerably. Using the decennial censuses acquired from the National Historical Geographic Information System (NHGIS) , racial and ethnic categories were simplified to 'white,' 'black,' and 'other,' with Hispanic/Latino being considered 'other.' This was done to allow for longitudinal comparisons across the decades, since the US Census Bureau's categorization of racial and ethnic identity has been inconsistent over time.
Results of quantitative and qualitative analyses
Even though the OLS statistics suggest a trend toward racial/ethnic minorities and lower SES populations having higher access to parks and physical activity sites, the GWR points towards "unpatterned inequity," meaning that disparities in park access exist, but the inequity is inconsistently correlated with specific socio-demographic variables. This is further supported by the qualitative analysis, which implies that a global (city-wide) analysis of accessibility may not be the appropriate analytic method for this data. A limitation of the qualitative analysis is that only two case study park areas were examined, which is not likely to be representative of all the parks in New York City.
The variability of the quantitative and the qualitative results suggests a number of potential limitations and shortcomings of our analyses. These limitations are discussed in detail below.
As noted earlier in this paper, there are many pitfalls and problems in developing a meaningful park accessibility measure, and our analyses have reinforced the need for a more comprehensive approach. A number of additional variables might be included when measuring park access, in order to potentially establish more definitive results. In addition, our analysis focused on park access in New York City, which may or may not conform to the realities of other geographies.
Proximity analysis based on "Walkability" distances
Access based on proximity or distance often does not take into account the actual street network, as noted earlier, and there may be a major highway or other barrier between the residents and the park. This was shown in our qualitative analysis of Highland Park, which is essentially cut off from the residential neighborhood on the northern side by large stretches of cemeteries and highways. Utilizing a network analysis may prove to be more realistic than the kernel estimation that we used, that quantified park access as a function of density of park acres and/or physical activities sites.
Actual points of entry to the parks
A measure of access also needs to consider actual points of entry to the park. For some parks, there are entry points that may be at a far remove from the residential neighborhoods, even though the park may border the neighborhood. This can also be addressed using network analysis provided that the entry points are known and mapped.
Most accessibility measures do not take into account perceptual access. For example, racial/ethnic minority residents might not feel welcome in a park or recreational facility used predominantly by non-Hispanic white individuals. Similarly, female park users might not feel welcome in male-dominated physical activity sites such as basketball courts.
Perceptual access can also be based on park cleanliness or perceived or actual crime within or near the park. A park may be in close proximity but unused due to the bad conditions within the park. These "incivilities," such as graffiti, broken glass, liter, or evidence of drug or alcohol use, or the presence of violent crime could be important factors in making the park unattractive for use. While it is difficult to include perceptual access into a measure or index, if data on park maintenance and crime rates are available, they could be included in an analysis. Additionally, other more qualitative methods, such as interviews and surveys of local residents, as well as cognitive (mental) mapping and participatory GIS, can be used to discern attitudes and perceptions about access to parks and physical activity sites.
Most measures of accessibility do not consider park characteristics, either, such as types and quantities of activities available, or park size. A tiny vest-pocket park will not have as much to offer in the way of physical activity potential as a large park. Although our study did incorporate park acres and number of activity sites, it could be improved by looking at variation of the types of physical activity sites, (e.g., does the park contain only basketball courts, or are there tennis, basketball, and a variety of other types of sites within a given park?). Presumably, a park having a greater variety of types of physical activity sites would make the park more of a draw to residents of different ages, genders, and physical fitness status, and therefore that park would merit a higher rating.
Another park characteristic that may be taken into account is the requirement of permits in some physical activity sites. For instance, permits are required to use the tennis courts in all parks. Since these permits have fees, access is limited based on financial ability.
Other variables to be considered
Resources available in parks and physical activity sites tend to be team sports-related, making it is less likely that older adults make use of these facilities. Therefore, the results of using physical activity sites as an important metric of accessibility may be misleading and skewed toward younger populations.
Additionally, people living in suburban-like areas within the city may have access to private open space, like backyards, usually available in single-family home neighborhoods, but not in higher-density, inner-city communities. Therefore proximity to public parks may be less important in those suburban-type areas, making any direct comparison of park and physical activity site accessibility between various types of neighborhoods and populations inaccurate. Public parks may serve a more critical function and there may be a higher need for public open space in less affluent neighborhoods, so statistical measures of equity regarding park access may not tell the whole story.
A valuable data source that was not explored in our qualitative analysis was comprehensive interviews with residents of the study areas to better understand actual and perceptual park accessibility. This could potentially provide important information necessary to address many of the limitations mentioned above.
A major limitation of population studies such as this one is the necessarily heavy reliance on data from the census. While census data is the most complete and current dataset we have at any given time, we need to acknowledge several underlying problems with its accuracy. One of the most serious sources of inaccuracy is the potential for undercounting populations in poor and immigrant communities. This has been an on-going drawback throughout the United States, but is even more pronounced in 21st century New York City, where a very high proportion of the population is foreign-born, and less likely to be counted in the census, especially in the case of illegal immigrants who may be mistrustful of government and wish to remain unknown to them. The temporarily or permanently homeless also comprise a significant population that is traditionally undercounted, as well as populations who may rotate their domicile and are therefore often overlooked in the official count because they are not thought of as being a permanent part of the respondent's household.
Additionally, each decennial census defines racial and ethnic categories differently, making cross-census comparisons difficult for longitudinal studies. The guidelines and standards for racial and ethnic classification were revised by the Office of Management and Budget in 1999, and the 2000 census uses a markedly different classification system from the previous censuses, making the findings of longitudinal studies somewhat unreliable.
It is generally acknowledged that access to parks and physical activity sites has beneficial health ramifications, so a better understanding of which populations have good access to these areas will assist in identifying and targeting those areas that do not, and the potential for more fully explaining disparities in health outcomes. It is important to recognize the environmental justice implications of park and physical activity site location and spatial distribution, since ethnic and racial minorities and poor people tend to suffer disproportionately from diseases which are often preventable by proper exercise.
While we did not find an overall environmental justice impact for New York City as a whole with regard to park access and socio-demographic indicators, we know that there are many sections of the city with poor access to parks, and therefore this needs to be examined on a very local level rather than globally.
The level of need also has to be taken into account because even if parks were distributed evenly throughout the city, some neighborhoods warrant having additional resources. This may be due to the fact that these neighborhoods are more densely settled than other more suburban-like parts of the city, and their populations do not likely have additional open space resources such as backyards or options to leave the city for recreational opportunities. These are the very populations for whom parks assume an even more critical function than typically provided, and the parks and recreational facilities in these areas perhaps deserve extra resources and funding commitments in order to provide the equivalent level of support.
Physical, cultural, and perceptual barriers should be taken into account also when measuring access to parks. Even though distances may appear short, true access cannot be gauged through Euclidean measures, and much more research has to be done on what constitutes true access and equity of resources.
The combination of quantitative and the qualitative analyses revealed trends that may not have been otherwise detectable. While the OLS (global) regression showed a weak relationship between SES characteristics and park accessibility, the geographically weighted regression (local) found "unpatterned inequality." The qualitative analysis did not reveal anything that would refute the statistical findings of the GWR analysis, as both neighborhoods were confirmed to be different from one another in terms of SES characteristics, although the park conditions and useable park area were comparable. The qualitative analysis did, however, suggest that an approach which considers physical barriers and some of the other variables listed in the limitation section would improve the model and better reflect reality. The qualitative analysis also showed that although the demographics around the study areas were similar at the times of the parks' construction, they have changed considerably since then, allowing for the possibility of environmental justice impacts. These environmental justice impacts may introduce disparities by influencing health outcomes and behaviors.
The complexity of the issues around racial and ethnic disparities in park access has been demonstrated further by this study. Looking at one factor at a time is likely to result in misleading findings. Therefore, a more complex model that accounts for as many different types of variables as possible (park size, access points, barriers, network distance, perception of safety, crime rates, park maintenance, availability and variation of physical activity sites) will be needed to develop a more accurate measurement of park accessibility, particularly as to how it might mediate environmental justice and mitigate negative health outcomes.
This research was partially supported by grant P60-MD0005-03 from the National Center for Minority Health and Health Disparities, National Institutes of Health. Thanks also are due to the New York City Department of Parks and Recreation for working with the student interns and Prof. Maantay at Lehman College's Urban GISc Lab to collect and map the park data in the comprehensive New York City-wide "Geographic Feature Identification Project."
- Boer J, Pastor M, Sadd J, Synder L: Is there environmental racism? The demographics of hazardous waste in Los Angeles County. Social Science Quarterly. 1997, 78 (4): 793-810.Google Scholar
- Bowen WM, Salling MJ, Haynes KE, Cyran EJ: Towards environmental justice: spatial equity in Ohio and Cleveland. Annals of the Association of American Geographers. 1997, 85 (4): 641-663. 10.1111/j.1467-8306.1995.tb01818.x.View ArticleGoogle Scholar
- Burke L: Race and environmental equity: a geographic analysis in Los Angeles. Geo Info Systems. 1993, 44-50.Google Scholar
- Chakraborty J, Armstrong M: Exploring the use of buffer analysis for the identification of impacted areas in environmental equity assessment. Cartography and Geographic Information Systems. 1997, 24 (3): 145-157. 10.1559/152304097782476951.View ArticleGoogle Scholar
- Chakraborty J, Schweitzer L, Forkenbrock D: Using GIS to assess the environmental justice consequences of transportation system changes. Transactions in GIS. 1999, 3 (3): 239-258. 10.1111/1467-9671.00020.View ArticleGoogle Scholar
- Maantay J: Mapping Environmental Injustices: Pitfalls and Potential of Geographic Information Systems in Assessing Environmental Health and Equity. Environmental Health Perspectives. 2002, 110 (2): 161-171.PubMedPubMed CentralView ArticleGoogle Scholar
- Morello-Frosch R, Pastor M, Sadd J: Environmental justice and southern California's "riskscape" – the distribution of air toxics exposures and health risks among diverse communities. Urban Affairs Review. 2001, 36 (4): 551-578. 10.1177/10780870122184993.View ArticleGoogle Scholar
- Neumann CM, Forman DL, Rothlein JE: Hazard screening of chemical releases and environmental equity analysis of populations proximate to toxic release inventory facilities in Oregon. Environmental Health Perspectives. 1998, 106 (4): 217-226. 10.2307/3433967.PubMedPubMed CentralView ArticleGoogle Scholar
- Perlin SA, Setzer RW, Creason J, Sexton K: Distribution of industrial air emissions by income and race in the United States: an approach using the toxic release inventory. Environmental Science Technology. 1995, 29 (1): 69-80. 10.1021/es00001a008.PubMedView ArticleGoogle Scholar
- Pollock PH, Vittas ME: Who bears the burden of environmental pollution? Race, ethnicity, and environmental equity in Florida. Social Science Quarterly. 1995, 76 (2): 294-309.Google Scholar
- Sheppard E, Leitner H, McMaster R, Tian H: GIS-based measures of environmental equity: Exploring their sensitivity and significance. Journal of Exposure Analysis and Environmental Epidemiology. 1999, 9: 18-28. 10.1038/sj.jea.7500023.PubMedView ArticleGoogle Scholar
- Bryant B, ed: Environmental Justice: Issues, Policies, and Solutions. 1995, Washington: Island PressGoogle Scholar
- Bullard R, ed: Unequal protection: environmental justice and communities of color. 1994, San Francisco: Sierra Book ClubGoogle Scholar
- Johnston B, ed: Who Pays the Price? The Sociocultural Context of Environmental Crisis. 1994, Washington, DC: Island PressGoogle Scholar
- Ball K, Bauman A, Leslie E, Owen N: Perceived environmental aesthetics and convenience and company are associated with walking for exercise among Australian adults. Prev Med. 2001, 33: 434-440. 10.1006/pmed.2001.0912.PubMedView ArticleGoogle Scholar
- Diez Roux A, Evenson K, McGinn A, Brown D, Moore L, Brines S, Jacobs D: Availability of recreational resources and physical activity in adults. American Journal of Public Health. 2007, 97: 493-399. 10.2105/AJPH.2006.087734.PubMedView ArticleGoogle Scholar
- Duncan M, Mummery K: Psychosocial and environmental factors associated with physical activity among city dwellers in regional Queensland. Prev Med. 2005, 40: 363-372. 10.1016/j.ypmed.2004.06.017.PubMedView ArticleGoogle Scholar
- Giles-Corti B, Broomhall M, Knuiman M, Collins C, Douglas K, Ng K, Lange A, Donovan R: Increasing walking: How important is distance to, attractiveness, and size of public open space?. American Journal of Preventive Medicine. 2005, 28 (2S2): 169-176. 10.1016/j.amepre.2004.10.018.PubMedView ArticleGoogle Scholar
- Wendel-Vos G, Schuit A, de Niet R, Boshuizen H, Saris W, Kromhout D: Factors of the physical environment associated with walking and bicycling. Med Sci Sports Exerc. 2004, 36: 725-730. 10.1249/01.MSS.0000121955.03461.0A.PubMedView ArticleGoogle Scholar
- Bell J, Wilson J, Liu G: Neighborhood greenness and 2-year changes in body mass index of children and youth. American Journal of Preventive Medicine. 2008, 35 (6): 547-553. 10.1016/j.amepre.2008.07.006.PubMedPubMed CentralView ArticleGoogle Scholar
- Berke E, Koepsell T, Moudon A: Association of the built environment with physical activity and obesity in older persons. American Journal of Public Health. 2007, 97 (3): 486-492. 10.2105/AJPH.2006.085837.PubMedPubMed CentralView ArticleGoogle Scholar
- Frank L, Andersen M, Schmid T: Obesity relationship with community design, physical activity and time spent in cars. American Journal of Preventive Medicine. 2004, 27 (2): 87-96. 10.1016/j.amepre.2004.04.011.PubMedView ArticleGoogle Scholar
- Gordon-Larsen P, Nelson M, Page P, Popkin B: Inequality in the built environment underlies key health disparities in physical activity and obesity. Pediatrics. 2006, 117: 417-424. 10.1542/peds.2005-0058.PubMedView ArticleGoogle Scholar
- Papas M, Alberg A, Ewing R, Helzlsouer K, Gary T, Klassen A: The built environment and obesity. Epidemiologic Reviews. 2007, 207 (29): 129-143. 10.1093/epirev/mxm009.View ArticleGoogle Scholar
- Taylor W, Poston W, Jones L, Kraft M: Environmental justice: Obesity physical activity and healthy eating. Journal of Physical Activity and Health. 2006, 3 (suppl1): 30-54.Google Scholar
- Rundle A, Field S, Park Y, Freeman L, Weiss C, Neckerman K: Personal and neighborhood socioeconomic status and indices of neighborhood walk-ability predict body mass index in New York City. Soc Sci Med. 2008, 67 (12): 1951-1958. 10.1016/j.socscimed.2008.09.036.PubMedPubMed CentralView ArticleGoogle Scholar
- Lantz P, House J, Lepkowski J, Williams D, Mero R, Chen J: Socioeconomic Factors, Health Behaviors, and Mortality Results from a Nationally Representative Prospective Study of US Adults. Journal of the American Medical Association. 1998, 279: 1703-1708. 10.1001/jama.279.21.1703.PubMedView ArticleGoogle Scholar
- Sorlie P, Backlund E, Keller J: US mortality by economic, demographic, and social characteristics: the National Longitudinal Mortality Study. American Journal of Public Health. 1995, 85 (7): 949-956. 10.2105/AJPH.85.7.949.PubMedPubMed CentralView ArticleGoogle Scholar
- Estabrooks P, Lee R, Gyurcsik N: Resources for physical activity participation: Does availability and accessibility differ by neighborhood socioeconomic status?. Annals of Behavioral Medicine. 2003, 25: 100-104. 10.1207/S15324796ABM2502_05.PubMedView ArticleGoogle Scholar
- Timpiero A, Ball K, Salmon J, Roberts R, Crawford D: Is availability of public open space equitable across areas?. Health Place. 2007, 13 (2): 335-340. 10.1016/j.healthplace.2006.02.003.View ArticleGoogle Scholar
- Abercrombie L, Sallis J, Conway T, Frank L, Saelens B, Chapman J: Income and racial disparities in access to public parks and private recreation facilities. American Journal of Preventive Medicine. 2008, 34 (1): 9-10.1016/j.amepre.2007.09.030.PubMedView ArticleGoogle Scholar
- Talen E: The social equality of urban service distribution an exploration of park access in Pueblo Colorado and Macon Georgia. Urban Geography. 1997, 18 (6): 521-541.View ArticleGoogle Scholar
- Wolch J, Wilson J, Fehrenback J: Parks and park funding in Los Angeles: an equity mapping analysis. Urban Geography. 2005, 25: 4-35. 10.2747/0272-36184.108.40.206.View ArticleGoogle Scholar
- Nicholls S: Measuring the accessibility and equity of public parks: A case study using GIS. Managing Leisure. 2001, 6: 201-219. 10.1080/13606710110084651.View ArticleGoogle Scholar
- Talen E, Anselin L: Assessing spatial equity: An evaluation of measures of accessibility to public playgrounds. Environment and Planning A. 1998, 30: 595-613. 10.1068/a300595.View ArticleGoogle Scholar
- Levine N: CrimeStat III: A Spatial Statistics Program for the Analysis of Crime Incident Locations (version 3.0). 2004, Washington DC: Ned Levine & Associates, 8: 1-Google Scholar
- Gatrell A, Bailey T, Diggle P, Rowlingson B: Spatial point pattern analysis and its application in geographical epidemiology. Transactions of the Institute of British Geographers. 1996, 21: 256-274. 10.2307/622936.View ArticleGoogle Scholar
- Herrmann C, Maroko AR: Crime Pattern Analysis: Exploring Bronx Auto Thefts using GIS. GIS for the Urban Environment. Edited by: Maantay JA, Ziegler J. 2006, Redlands, CA: Environmental Systems Research Institute (ESRI), 407-413.Google Scholar
- Moore L, Diez Roux A, Evenson K, McGinn A, Brines S: Availability of recreational resources in minority and low SES areas. American Journal of Preventive Medicine. 2008, 34 (1): 16-22. 10.1016/j.amepre.2007.09.021.PubMedPubMed CentralView ArticleGoogle Scholar
- Fotheringham AS, Brunsdon C, Charlton M: Geographically weighted Regression: the analysis of spatially varying relationships. 2002, West Sussex, England: John Wiley & Sons LtdGoogle Scholar
- New York City Department of Parks & Recreation: Highland Park.http://www.nycgovparks.org/parks/highlandpark
- New York City Department of Parks & Recreation: Marine Park.http://www.nycgovparks.org/parks/marinepark
- National Historical Geographic Information System (NHGIS).http://www.nhgis.org/
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.