Back to Table of Contents

Abundant Fields, Meager Shelter:
Findings From a Survey of Farmworker Housing in the
Eastern Migrant Stream

© Housing Assistance Council, 2000

Permission is granted ONLY to nonprofit community-based organizations to reproduce and/or adapt this document, and only for their own use.

APPENDIX D: COLLECTION OF INFORMATION EMPLOYING NONPROBABILITY SAMPLING

This section discusses the strengths and limitations of the survey methodology, and explains how the methodology was designed to address the most significant possible limitations in the data collection. Since surveys were completed as FHSI staff performed their healthcare outreach duties, the data may be characterized as a convenience sample. This means that the data collection was not entirely random, although randomness is desirable when making statistical generalizations and inferences from the data set. A nonrandom convenience sample therefore results in a nonprobability sample, which has different types of potential bias than a statistically random sample.

Limiting Bias and Maximizing Response Rates of the Nonprobability Sample

While probability sampling is preferable in survey work, collecting a nonprobability sample is often necessary due to the nature of the population under study. Nonprobability sampling is often the only recourse for researchers interested in gathering information on populations that are difficult to define, locate or recruit. Because migrant and seasonal farmworkers are mobile, and many do not speak English, researchers and government have had difficulty gathering accurate demographic and housing data. A nonprobability sampling methodology as used in this study represents one of the only feasible options for gathering accurate information on farmworker housing conditions.

Nonprobability sampling has some limitations in generalizing from the data obtained, and is subject to a number of areas of potential bias in the data collection process. This study was designed to gather data that is representative of housing conditions in regions of the Eastern migrant stream with the heaviest concentrations of migrant and seasonal farmworkers. The design of this study also controlled for the most significant sources of bias in data collection.

One limitation was that the survey of units upstream would not be as concentrated as the survey of units in Florida. The greater concentration of surveys in Florida was a necessary byproduct of FHSI staff schedules and migrant patterns. While it was anticipated that the same number of surveys would be completed upstream as in Florida, the upstream surveys would capture information from a wider geographical area with more widely varying housing and economic environments. Although this places some limitation on generalizations from the upstream data beyond the regional level, this limitation is offset by the uniqueness and quality of the data set. This information will retain significant value as a selection of regional "snapshots" of East Coast farmworker housing conditions. As the data collection is pursued in upcoming years, and expanded to the Midwest and West Coast migrant streams, stronger observations may be made concerning regional and national patterns evident in farmworker housing conditions.

There are a number of areas in which bias may be introduced in the collection of a nonprobability sample. Potential sources of bias include the inexperience of FHSI staff in conducting housing assessments, the enlistment of cooperation among respondents, the availability of respondents, miscommunication in survey response, and a skewed representation of unit types among completed surveys.

Some error may be introduced into the data collection because of the inexperience of FHSI staff with housing assessment. HAC addressed this potential source of error through training prior to commencement of survey work. FHSI staff do, moreover, have extensive experience in survey-related work as part of their healthcare outreach. Skills such as developing rapport, posing questions simply and clearly, and reaching a wide range of respondents in a given area are necessary for the performance of the healthcare outreach duties of FHSI staff.

HAC established limited quota guidelines to ensure that the range of housing structure types occupied by farmworkers in a given region are represented. At many farm labor camps, there is often a mix of housing structures. For example, a farm labor camp may have a dormitory for housing single migrants and a number of small single-family, detached units for use by families. FHSI staff were required to survey at least one unit of each structure type present at a given site. In multi-unit structures (such as an apartment building or a dormitory), FHSI staff were asked to survey at least one unit of each bedroom size in a structure with different unit sizes. FHSI staff were also required to survey at least two units in each multi-unit structure, even if all of the units in the multi-unit structure are of the same size. Since housing cost burden and crowding calculations are sensitive to the size of a housing unit and family size, these efforts should account for some of the range of housing costs that may be faced by households or individuals living in the same apartment complex or dormitory.

The failure to enlist the cooperation of respondents presents another opportunity for bias in the data collection. If many potential respondents do not want to participate in the survey, then significant subgroups may be underrepresented in the sample. However, FHSI staff have a positive reputation in the farmworker communities in which they work, and have been trained in developing rapport with farmworker households. Since many farmworkers do not speak English, or speak English as a second language, miscommunication may contribute to nonresponse. FHSI staff are bilingual in English and Spanish, which reduces the incidence of distorted responses when surveying individuals who do not speak English or for whom English is a second language. Many FHSI staff also have skill in other languages prominent in the East Coast migrant stream, such as Haitian Creole. This language proficiency also promotes the development of rapport between outreach workers and the farmworker respondents. These aspects of FHSI job skills reduce the potential of bias from failing to enlist the cooperation of respondents or miscommunication.

Another source of sampling bias concerns the issue of availability. For example, biased data may result if a survey is taken on weekdays during business hours, so that the only individuals likely to respond are those who do not work. Healthcare outreach takes FHSI staff to different locations at different times of the day. They meet farmworkers at the work site during the day, and they also meet with farmworkers during weekends and evening hours in their homes and at public gathering places such as laundromats. This pattern of outreach among FHSI staff meant that availability of respondents would not pose a problem in assessing housing units. Their work hours performing outreach ensured that they visit a significant number of housing units at different times of the day, which improved the likelihood they would be able to assess unit interiors and survey residents, or set up housing assessment appointments while doing outreach at work sites or public gathering places.

Because this data collection was performed as a convenience sample, some types of units may receive more representation in some survey areas. This is a natural outgrowth of FHSI outreach efforts, where they must target their services to areas with the greatest concentration of farmworkers. In some counties this may mean that FHSI services are primarily being utilized in farm labor camps, and in others it may be the case that more outreach is being performed among households residing in private market housing. In these cases, one type of housing situation will receive more representation than other types. HAC addressed this potential source of sampling bias through its training of FHSI staff by providing instruction on how to maximize the range of data collected without significantly detracting from FHSI staff commitments to perform their healthcare work. Additionally, most FHSI staff have knowledge of the areas to which they are assigned. These staff have expertise in the range of housing options available to farmworkers in each service area. Finally, FHSI outreach includes health education and informational services, and their staff do not work exclusively at camps and housing units in great disrepair. In the pretest of the survey instrument, a number of surveys returned indicated good quality dormitories and apartment units at the sites visited. While there is no way to eliminate the potential for this kind of sampling bias completely, HAC's training and FHSI expertise in the field should limit the overrepresentation of particular types of units in different regions of the East Coast migrant stream.

Test of the Survey Instrument and Methodology

HAC and FHSI performed a pretest with a preliminary draft of the survey instrument during the summer growing season of 1997. FHSI staff performed surveys at two sites in Virginia, one site in Tennessee, and one in Massachusetts. Surveys contained information on 109 housing units. Experience in administering the survey during the pretest was used to revise the survey instrument for clarity and brevity. Additionally, experience gained in pretesting the instrument was used to develop training for FHSI staff who participated in survey work in FY1998.

On To Appendix E

Back To Table of Contents