Section VI - Data Analysis
Overview
The information collected in each of the counties was sent to the Office of Research and Statistics for data entry and analysis. In addition, the Office of Research and Statistics received HMIS data from four of the five continua of care. A data analysis team consisting of staff at ORS and representatives with the South Carolina Council on Homelessness met periodically during this period to resolve questions with the forms and any ambiguities with the data. If necessary, the county coordinator or local agency was contacted for clarification. Data entry was carefully monitored to ensure consistency in resolving issues as well as data quality. Forms were color coded to aid data entry. In addition, each form was given a unique number so that if issues did arise, it would be possible to review the actual form.
Issues were found during the data entry process. While the forms were pre-tested with a limited population in the Columbia area, several issues arose on the instruments themselves. For example, an option for "refused" and "don't know" probably should have been added to many of the questions. Simple instructions such as "continued on next page" should have been added at the bottom of each page for the multi-page forms. Many volunteers only filled out the first page on double-sided forms. On the sheltered form, checkboxes for "Emergency" or "Transitional" needed to be added. The address where the forms should be sent should have been on the form itself and would have saved some confusion. There appeared to be some confusion on who should use the family forms - leading to the need to emphasize that a family form should be used only if children under 18 are present. Occasionally the list of options under some of the questions was too lengthy. For the question "Where did you stay tonight" - the list may have been too long. It appeared that the volunteers rather than reading through all the options, tended to group the answer under "Other". The entire issue on whom and what constituted homelessness led to confusion. One way to possibly address this issue is to provide a list of which situation do / do not constitute homelessness with each survey. Because the process used volunteers and service providers who were no doubt time-strapped, there were considerable problems with handwriting issues. Some forms were completely illegible and therefore were not usable.
In analyzing and examining the HMIS data from the four continua of care, a number of data issues arose. While a list of critical data elements was created to guide the HMIS database consultant on what elements to extract, not all the requested data elements were extracted from HMIS. No standardized layout was used in the extraction process. In addition, after closer examination and discussions with the local continua database consultants, in some instances the data elements varied slightly in content and/or definition. This led in some instances to the need for new extracts to be created and subsequently re-analyzed.
Once all of the data entry forms were cleaned and then entered, databases had to be created and un-duplicated. This was a multi-step process and presented a number of challenges. Databases were created for each form and for each of the HMIS systems. Within each database, duplicates were eliminated primarily based on the unique combination of identifiers created. However, great care was taken to ensure that observations that appeared to be duplicates were truly duplicates. In some instances, the actual forms were examined to ensure that the observation was a duplicate. Because only a limited number of identifiers were requested and in some instances, not all the identifiers were actually submitted - this cautionary step was deemed important. It was also necessary to ensure that fieldnames were identical across the databases. Finally, all the databases were appended and un-duplicated across the databases.
The questions on the survey allowed the Office of Research and Statistics to determine how many of the individuals who completed the survey or were in the HMIS system were homeless on January 25th. In addition, ORS was able to provide additional detail on key homeless sub-populations. A list of definitions for these sub-populations can be found in Appendix C.
Results from the 2007 Homeless Count
The 2007 Homeless Count found a total of 6,759 homeless individuals in South Carolina. Unfortunately, these results should not be compared to the results of the 2005 Homeless Count. The methodologies used in the 2007 Count to the 2005 are different - care must be taken in any comparisons. While it is thought that the 2007 Homeless Count was a "better" count (better organization, better methodologies, etc) - in truth - it is extremely difficult to know for a certainty. Variables as simple as the weather on the actual night of the count or the number of volunteers who actually come out to help can impact the count. Therefore it is important to think of the homeless count as more of a "moving target" - literally and figuratively.
Those 6,759 individuals represent .16% of the total population in the state and 1.1% of the persons below poverty (according to the 2004 SAIPE estimate by the US Bureau of the Census. Of the 6,759 individuals who were found homeless, 5,594 were identified as homeless using HUD's strict definition (over 80%) leaving 1,165 identified as homeless through a broader definition. Over 50% of the homeless population was found in only five counties: Greenville, Richland, Horry, Georgetown, and Charleston - all who were considered to have a high level of effort. In addition, Greenville, Richland, and Charleston are highly urban. While Horry and Georgetown are perhaps not considered to be as urban as Greenville, Richland, or Charleston - Horry and Georgetown counties represent a unique part of South Carolina's geography - the coast with high transitional populations. Over 75% of the homeless population were in 11 counties (out of the 46 total counties in South Carolina) - again all with a moderate to high level of effort with the exception of Clarendon. These remaining counties (Greenwood, York, Lancaster, Clarendon, Anderson, and Spartanburg) are urban to somewhat urban (except for Clarendon) - thereby more likely to be offering services for the homeless populations. Clarendon County was one of the four counties who used the observation methodology and subsequently utilized "Form B" which collected information about the number of homeless people observed. That form had no identifiers and therefore it was not possible for any un-duplication.
Of the 6,759 homeless individuals, 2,996 (44%) were found in shelters. Those individuals not found in shelters were either identified through the street counts or through the identification of some of the other homeless populations such as the "doubled up". As has been denoted before, identification of the homeless population outside of shelters is a very difficult task and almost by definition would be an undercount. Of the sheltered population, 1,650 (55%) were found in emergency shelters with the remaining 45% (1,346) in transitional shelters.
The faces of homelessness are often different from the perceptions by the general public. While over 80% (5,430) were adults, close to 20% (1,329) were children (ages less than 18 years of age). Some of these children were in families but 33 of these children were considered to be unaccompanied youth. Over one third (31%) of the homeless population were in families with dependent children. Slightly higher proportions (38%) were found in the more extended definition of the "other homeless population". That is not surprising given that a broader definition which included situations such as temporarily staying with families or friends also commonly known as "doubling up" was used.
Within the homeless population, there are other key sub-populations. One sub-population of particular interest to HUD is the chronically homeless which were 477 individuals representing 7% of the total homeless. Chronic Homelessness refers to an unaccompanied homeless individual with a disabling condition who has either been continuously homeless for a year or more OR has had at least four (4) episodes of homeless in the past three (3) years. To be considered chronically homeless, persons must have been sleeping in a place not meant for human habitation and/or in an emergency shelter during that time.
Fifteen percent (15%) of the homeless were considered to have a chronic substance abuse issue. Slightly over 7% were severely mentally ill. (These categories overlap and therefore should not be added together - but of course - many homeless have co-occurring issues. There was a slightly higher proportion of the severely mentally ill identified in the "Other homeless" population (9%). It is important to note that some of the data on sub populations was collected through interviews in which people self reported their conditions.
Additional key sub-populations were veterans at almost 8% or 517 individuals and persons identified with HIV/AIDS with 1% of the population. Victims of domestic violence are defined as persons who have fled housing or might flee housing as a result of emotional or physical abuse at the hands of a spouse, minor child or parent (if minor child). They presented almost 9% of the total homeless population (or 594 individuals). As a proportion of the "other homeless," victims of domestic violence represented close to 13% of that population. Again, that is not surprising given the addition of "doubling up" used in the broader definition.
The State and County Data Tables provide additional detail.
Comparisons of the 2007 Homeless Count to Other Estimates
Because of the many challenges in counting the homeless, this report offers additional estimates of the Homeless population for comparisons.
One common methodology to derive additional point-in-time (PIT) estimates is to base the annualized estimate on the latest poverty estimates. Literature suggests that people who are homeless represent approximately 6.3% of the persons considered to be below poverty. Using the 2004 Census Bureau's SAIPE estimates by county (latest available), annualized estimates were first derived by multiplying 6.3% to the Census Bureau's poverty estimate. Literature suggests that the homeless annualized estimate ranges from 3 to 6 times higher than a point-in-time. Therefore, the annualized homeless estimate derived from the poverty estimate was converted to a PIT by dividing the estimate either by 3 or 6. (Both estimated PITs were calculated.) Because poverty estimates were available by counties, additional estimates were calculated at the county level.
Based on the poverty estimating methodology, the PIT estimates ranged from almost 13,000 to 6,500 (compared to the 6,759 Homeless individuals found in the 2007 Count). On the surface, this range would indicate that the number found in the homeless count was a reasonable count though definitely at the lower end of the range. However, given South Carolina's demographic dynamics such as a "housing stock" that is still below standards in some areas of the state, a population mired in persistent poverty, and a highly transitional population in some coastal areas; one would suspect that the 6,759 individuals found represents an undercount.
The PIT estimates at the county level provide additional speculation. For those counties with none to a very low level of effort; the difference in the total homeless count to the estimated PITs is dramatic - with the total homeless count appearing to be a serious undercount (with the exception of Clarendon). The difference is less dramatic for those counties with a low to moderate level of effort. For those counties labeled with a moderate to high level of effort, the picture is mixed. For several counties, their homeless count clearly falls in the range between the 2 PITs based on poverty. However, there were a few counties whose count still fell below even the lowest range point for the 2 PITs raising the speculation that perhaps the level of effort was more moderate than high. Four counties had counts higher than the higher range point for the 2 PITs - Georgetown, Greenwood, Horry, and Lancaster. One area of exploration would be to see if those counties utilized extraordinary efforts coupled with a methodology that allowed them to locate more of their homeless populations.
Annualized estimates provide another framework to understanding the homeless population. As discussed earlier, the homeless population tends to be highly transitional with "cycles of homelessness". A point in time count is more likely to capture people who are homeless for long periods of time. Thus, people with the most obstacles to recovery from homelessness (e.g. mental illness, addiction) can be over-represented and those who typically experience shorter episodes, such as families, can be under counted. An annualized estimate helps providers to plan for services other than beds.
As mentioned earlier, one annualized estimate was developed utilizing the poverty estimates by county. However, another methodology actually incorporates the total PIT homeless count coupled with information on the cycles of poverty. The formula to convert a PIT homeless count to an annualized estimate is a follows:
A + ((B*51) * (1-C)) = Annual Estimate where:
A = PIT count of homeless people
B = the number of people who became homeless with the last 7 days, whether for the first time or not, or were already homeless but just entered the boundaries of your community within the past 7 days.
C= Proportion of homeless people who have had a previous homeless episode within the past 12 months
Unfortunately the information to calculate this estimate was only available using forms A and C. It was not available in form B nor was it available through the HMIS extracts. However, using the data available an estimate of 17,454 was constructed.
The following table provides additional detail:
a. 2007 Homeless Count Compared to other PIT Estimates and Annualized Estimates
Finally, one last set of estimates on homeless children will be included in this document. The U.S. Department of Education definition of homeless includes children and youth who are in shelters, lacking shelter or sharing the housing of other persons due to loss of housing, economic hardship, or a similar reason (sometimes referred to as doubled-up). The South Carolina Department of Education surveys each of its school districts using the above homeless definition - then aggregates the results for the state. The resulting tables are presented in Appendix D.