The Power of Zip Codes: An Analysis of FACT Location Data
by Dale E. Jones, Retired Research Director, Church of the Nazarene Global Ministry Center
The following article was written with social researchers in mind. However, there are implications for congregational leaders who are trying to understand how their local faith group relates to their community as well. Following each of the sociological paragraphs in normal type, there will be bold italicized commentary and insights designed for local congregations to consider.
Social researchers believe that community location may influence religious attitudes and practices. Churches in smaller communities may not need a robust on-line presence; synagogues in mid-size suburbs may require more off-street parking than those in central cities where mass transit is widely used.
But how does a researcher determine the appropriate community description?
Self-description of Location
In 2010, Faith Communities Today conducted a survey in which each faith community was identified by ZIP code. Each survey respondent was also given a chance to choose the best description of the community around the congregation.
Table 1: Self-described Locations of Participating Faith Communities
Descriptor | Responses | Percent |
Rural area or open country | 1,460 | 12.8% |
Village or town with a population of less than 10,000 | 2,401 | 21.1% |
Small city or large town with a population of 10K to 50K | 2,556 | 22.5% |
Downtown or central area of a large city | 912 | 8.0% |
Older residential area of a large city | 1,573 | 13.8% |
Older suburb around a large city with a population of 50K+ | 1,510 | 13.3% |
Newer suburb around a large city with a population of 50K+ | 682 | 6.0% |
None of these descriptors chosen | 279 | 2.5% |
But how well do those self-described locations reflect a standardized definition of community?
One ZIP code had five responses and each respondent chose a different place descriptor: Rural area or open country; Village or town with a population of less than 10,000; Small city or large town with a population of 10K to 50K; Older residential area of a large city; and, Older suburb around a large city with a population of 50K+. The zip code itself had a population of 39 thousand at the time of the survey.
The confusion is understandable. The 08844 ZIP code itself is in central New Jersey. The Census Bureau did not report any officially defined places (cities, towns, boroughs) or even unofficial places within its borders in 2010. In that sense, it could qualify as “open country.”
On the other hand, it borders Manville, Millstone, Raritan, and Somerville boroughs. Therefore, if one ignored corporate limits, it would be easy to consider oneself as part of a village of less than 10,000 people (near Millstone or Raritan) or a small city of 10,000 or more (Manville or Somerville).
However, most of the zip code is also included in the 18-million person New York-Newark, NY-NJ-CT urbanized area. If the respondents regarded themselves as part of greater New York, then they might indeed consider that they were in the residential area of a large city, or more specifically within an older suburb around a large city.
When analyzing survey responses, does one want an objective location definition for comparisons (“communities of faith in such defined locations have these characteristics”), or is the perception of the respondent more appropriate (“responders who consider themselves in such locations report that their faith communities have these characteristics”)? Either is a valid approach, but they are likely to yield disparate results.
How common is this disparity? In the 2010 survey, 7,596 different zip codes were linked with place-identified records. Of these, 1,372 (18.1%) had at least two different place descriptors in the results, and 198 had three or more. But those that had no discrepancy included 5,246 with only one responding congregation. That is, 69.1% of the respondents had no other respondents for possible comparisons. If only one respondent from ZIP code 08844 reported, there would have been no indication from the data that another identification of the location was possible.
Local congregations often perceive themselves in ways that do not match the reality of their neighborhoods. In the example above, five religious groups in the same zip code chose five different descriptions for their location. Later in the article, census data will be used to describe the community more objectively. But each congregation’s self-perception of the neighborhood probably speaks very much to the outreach methods employed by that group.
As the next section indicates, the self-perception can be affected by group history as well as by the actual demographic characteristics of the neighborhood. If a congregation has not done an objective analysis of its location, there will be a tendency to selectively interpret the information.
Standardized Location Description
To standardize location descriptions requires standardizing the definition of cities and towns. Eastern cities such as Boston or Philadelphia tend to be surrounded by highly developed suburbs that limit expansion of that central city. Southern and western cities such as Jacksonville or San Diego tend to have absorbed much of the surrounding territory.
In place of corporate limits, the urbanized areas and urban clusters as consistently defined by the U.S. Census Bureau can be used. If the center of the ZIP code is located within one mile of such an urban place, it will be initially classified by that community’s size.[1]
With the self-identification categories, well over half
(56.4%) of respondents were outside “large cities,” usually defined as having
at least 50 thousand people. In fact, nearly two-thirds (62.9%) of the ZIP
codes were centered within urbanized areas of at least that population. Including
those ZIPs that are centered within one mile of an urbanized area, thus
allowing for the larger geographic size of outlying ZIP territories, the ratio rises
to 65.6%.
Table 2: Standardized Locations of Participating Faith Communities Based on Population
Descriptor | Within 1 Mile | Percent |
Rural area or open country | 1,740 | 15.3% |
Village or town with a population of less than 10,000 | 970 | 8.5% |
Small city or large town with a population of 10K to 50K | 1,118 | 9.8% |
Larger city with of population of 50K+ | 7,457 | 65.6% |
None of these descriptors chosen | 88 | 0.8% |
Before considering how best to standardize “older”, “suburban,” and “residential,” it is worth exploring how these two approaches yield such different results.
The most important factor is definition of terms. As mentioned above, corporate boundaries can create a community that is a small town in its own right and is also a suburban enclave of a much larger community. And some communities may have the feel of a town though not actually within any corporate limits, leading different respondents to classify it as either a small city or as open country.
A survey instrument is not the best place to teach urban/rural classifications, nor to explain the difference between an urban cluster and an urbanized area. And a national survey can scarcely afford to define its questions in a way that recognizes the different community types within, for instance, Baltimore County, Maryland, and Los Angeles County, California. The latter has nearly all its citizens gathered into specific cities with corresponding city governments and officials. The former, despite having nearly one million people, did not contain a single incorporated city reported in the 2010 census. Yet each county is essentially part of a single urban agglomeration.
Again, it may be instructive to know that more than half of respondents classified themselves as part of smaller communities. This may indeed reflect the attitudes of people within their faith communities. But it would be very misleading to state that larger cities contain less than half of the nation’s worshipping centers.
Many surveys do rely upon self-identification for determining location data. The 2010 FACT survey allows us to see the potential for misunderstanding when that is the sole source of such information.
Especially in urban areas, most neighborhoods have a variety of applicable descriptors. The community may be technically within the boundaries of a specific city of a particular size; it may have its own central business district (whether composed of skyscrapers or a large mall); further, it may have some industry with employees who drive in from other communities; there also may be a significant number of commuters to other parts of the urban area; and it may still have a very old section of town along with newer housing developments.
The local congregation’s perception may be affected by its immediate neighborhood. If it’s located just off the business section of a small town now being engulfed by suburban development, it may still consider itself as downtown in its own city. If the congregation had moved to one of the new developing areas, it might consider itself part of a newer suburb.
The local congregation might instead be influenced by the self-perception of its constituents. If they are long-time residents of the same separate-city-becoming-suburb, they might see themselves as residents of a small or large city. Or, if those long-time residents are commuters, then they may consider themselves part of an older suburb despite the presence of newly built neighborhoods.
The important point is that many congregations’ self-description is more indicative of the group identification than the actual neighborhood.
Standardized Neighborhood Definitions
The four categories of large cities as defined in the FACT survey may not be the best descriptors available for census data. With census information tied to ZIP locations, it may be appropriate to base some comparisons upon ethnic diversity, age, education, and occupation or income categories. However, since the survey used the four categories shown above, an attempt has been made to define those terms based upon census information.
To determine whether a community is older, residential, and/or suburban can be determined by analyzing the age of housing, the type of housing, and the commuting times of census-defined block groups within one mile of the ZIP center.[2]
A high commuting ratio (HCR) was computed for each ZIP and each urbanized area based on the percentage of people who commuted at least 45 minutes to work. Any ZIP with a greater HCR than its urbanized area was considered to be suburban. However, this did not include many areas that are traditionally considered suburban, such as most of Central New Jersey. Therefore, any ZIP with at least 40% of its commuters traveling for at least 30 minutes was also categorized as a suburban location no matter the urban area HCR.[3]
If the suburban area housing stock included a higher ratio of units built since 2000 than its urbanized area, it was classified as a newer suburban area; otherwise, it was classified as an older suburban area.
The remaining ZIP codes were classified based on two criteria. Low commuting ratios (LCR) were calculated based on the percentage of people whose commute was less than 15 minutes. Housing density was computed based on the proportion of housing units with at least 10 units per address. Those ZIP codes with a 20% higher LCR than their urbanized area and with a 20% higher housing density than their urbanized area were classified as downtown or central city. The remaining ZIP codes were classified as residential.
The actual term used in the survey questionnaire was “older
residential area.” Some of the areas now classified as residential may in fact
contain a high proportion of new housing stock. Since the questionnaire seemed
to be distinguishing between suburban and central city more than between newer
and older communities, this refinement will automatically re-classify any
location where the respondent thought the more important distinction was age of
housing.
Based upon these classifications, each FACT survey can now be linked to a more consistently defined community type.
Table 3: Standardized Locations of Participating Faith Communities
Descriptor | Standardized | Percent |
Rural area or open country | 1,740 | 15.3% |
Village or town with a population of less than 10,000 | 970 | 8.5% |
Small city or large town with a population of 10K to 50K | 1,118 | 9.8% |
Downtown or central area of a large city | 1,550 | 13.6% |
Older residential area of a large city | 2,553 | 22.4% |
Older suburb around a large city with a population of 50K+ | 2,132 | 18.7% |
Newer suburb around a large city with a population of 50K+ | 1,222 | 10.7% |
None of these descriptors chosen | 88 | 0.8% |
As shown in Table 3, the largest number of responses came from neighborhoods that would be consistently described as “Older residential areas of a large city.” This ratio is 62% higher than the self-reported ratio in Table 1. In contrast, the largest groups in Table 1 were “Small city or large town with a population of 10K to 50K” and “Village or town with a population of less than 10,000.” Under a standardized definition, these two groups are each less than half as large and are now the least common descriptors.
For social researchers, the above categories offer one way of consistently defining a neighborhood. For local congregations, it is more important to grasp that the immediate community may have features (age of housing and commuting time in the above example) not readily apparent to those currently part of the congregation.
Comparison between Self-Identified Locations and Standardized Locations
Finally, what is the likelihood of self-identified locations matching the standardized definition?
Table 4 shows how well the self-identified locations in the FACT survey matched the standardized classifications identified above. This is not intended to criticize the individuals who responded, but to show the difficulty respondents have in providing community descriptions that are consistently applicable for statistical analysis.
Half of those who considered themselves in a rural area or open country were actually located in such areas, according to the population characteristics of their ZIP code. But one in five (20.4%) were actually located within, or at least adjacent to, an urbanized area with at least 50,000 people.
Since the standard cut-off for a village or town in the standardized approach was a minimum of 2,500 people, perhaps the majority of those who self-identified as village or town with less than 10,000 people selected an appropriate classification. But one in ten (10.2%) were part of urban clusters with at least ten thousand people and one in four (25.0%) were part of urbanized areas with at least fifty thousand.
Those who self-identified as being in a small city were nearly as likely to be in an older residential area of a much larger urban area. Over two-thirds (68.5%) of those who selected “city under 50K” were actually part of an urbanized area of at least fifty thousand.
Table 4: Likelihood of Self-identified Locations Matching Standardized Locations
Standardized | |||||||
Self-identified | Rural area | Village | Small City | Downtown or central | Older Residential | Older Suburb | Newer Suburb |
Rural area or open country | 52.1% | 14.8% | 12.0% | 2.7% | 7.1% | 4.3% | 6.2% |
Village or town with a population of less than 10,000 | 36.9% | 27.2% | 10.2% | 1.8% | 9.6% | 5.4% | 8.2% |
Small city or large town with a population of 10K to 50K | 1.7% | 3.1% | 25.9% | 13.0% | 23.8% | 17.2% | 14.5% |
Downtown or central area of a large city | 0.8% | 0.1% | 0.8% | 38.5% | 25.4% | 28.3% | 5.4% |
Older residential area of a large city | 0.4% | 0.3% | 0.3% | 24.3% | 32.9% | 35.0% | 6.5% |
Older suburb around a large city with a population of 50K+ | 0.7% | 0.1% | 0.1% | 16.9% | 37.6% | 32.5% | 11.7% |
Newer suburb around a large city with a population of 50K+ | 1.5% | 0.1% | 0.6% | 16.3% | 33.0% | 19.5% | 27.0% |
Each row adds to 100% of the responses listed in the left column. Thus, 52.1% of self-identified Rural area or open country respondents were found to be in such areas as standardly defined; 14.8% that chose that category were standardly classified as Village or town with a population of less than 10,000; and so on.
The distinctions within the large cities likewise differ. However, as noted above, the distinctions necessary for standardization are more difficult for casual observers to perceive. It is true that at least 98% of those selecting each of the large city categories were in fact located within such cities.
So how did our five-way zip code finally get classified? It is in an area where 41% of the population travels at least 30 minutes to work. That is above the 40% threshold, classifying it as suburban. Since 11% of its housing units were built in 2000 or later, which is more than twice the ratio for the urbanized area, it is classified as a newer suburban area. None of the five respondents chose that description. Since nearly nine of ten housing units were built before 2000, perhaps the “newer” definition could be challenged. But even with that change, only one of the respondents actually chose “older suburb.”
For social researchers, this section describes how inaccurate self-identification usually is. For local congregations, it is an encouragement to more fully study the actual community rather than rely on its own impressions.
Recommendation
When location information is available for survey respondents, usually based upon zip code or even specifically geo-coded addresses, it would normally be preferable to use that information to obtain community descriptions.
Some community distinctions are admittedly difficult to achieve through census information. (Endnote 3 mentions one difficulty in the system used for this paper, for instance.) But adding a few clarifying questions to the survey is still possible, while relying upon the location data to reflect many more community characteristics far more accurately.
Ideally, the location information should be obtained independently from the survey. Those supplying contact information should be asked to include sufficient data that at least a ZIP centroid could be used, if not an actual location address. This would require a password system to allow linking between the survey data and the faith community’s location.
Relying on a zip code supplied by the respondent would be acceptable, but this will increase the possibility of not being able to locate all surveys geographically.
If it is decided to rely upon self-identification of community type, the potential responses should be defined as clearly as possible. They should also be exhaustive and mutually exclusive, of course. In the survey above, for instance, would a newly restored section of south Chicago be considered “older residential area” or “newer suburb”? And the fact that a small city is also a suburban location is one of the reasons for inconsistent perceptions.
Local congregations desiring to know their own community better can use such sites as The Association of Religion Data Archives. Some denominations offer similar programs. The census bureau has the raw data, but tends to focus on city boundaries or larger areas rather than on smaller neighborhoods.
[1] Of 41,788 ZIP codes listed in the database for 2010 analysis, 15,677 were centered within urbanized areas (at least 50,000 population); 3,959 were centered within urban clusters (at least 2,500 population but less than 50,000); and another 2,745 were within one mile of an urbanized area or urban cluster. Of the latter, 36 were within a mile of two such places; these were assigned to the place considered by the researcher to be most closely linked by road networks.
[2] Several radii were considered, as was the composition of the ZIP code itself. The latter was impractical, as census data is not routinely updated by ZIP location. (ZIP boundaries are fluid. Even in this analysis, 88 respondents were reported to be in ZIP codes that were not included in the 2010 dataset.) Therefore, a standard radius was deemed to be more appropriate for determining the community type. The one-mile radius was thought to be large enough to describe the neighborhood without including too much beyond. Of all 156 thousand urbanized area block groups, 75% were within one mile of a ZIP centroid and 33% were included in more than one ZIP calculation. With a three-mile radius, 98% of block groups are included, but 92% were included in multiple ZIP calculations, greatly masking the differences between ZIP neighborhoods. At a half-mile radius, only 9% of urbanized area block groups were duplicated, but less than half (45%) were included in the calculations.
[3] Virtually all the ZIPs in Brooklyn and Queens have higher HCRs than the New York metro, while almost none in New Jersey exceed the urbanized area’s HCR. That is why the “or at least 40% traveling for half-an-hour” was added. The threshold could be set at another percentage, of course, but the principle behind the standardization remains the same.