Human Trafficking Policy and Research Analyses Project, Houston, Texas, 2020-2024 (ICPSR 39250)

Time Period(s) View help for Time Period(s)

Date of Collection View help for Date of Collection

Data Collection Notes View help for Data Collection Notes

Study Purpose View help for Study Purpose

The purpose of the present study was to examine the prevalence, risk factors, and protective factors of labor exploitation and trafficking among construction workers.

Sample View help for Sample

Study participants were sampled and recruited through two strategies: Time-locations sampling (TLS) and link tracing sampling (LTS). The goal was to generate separate prevalence estimates from each of these samples. The rest of this section details the original sampling plans and challenges that were encountered in the field.

TLS

TLS involves developing a sampling frame of venues, days, and times where the target population congregates and using a random selection procedure (e.g., every fifth person) to select a representative sample of the population. This was deemed a promising method to use among construction workers because they congregate at worksites at predictable time intervals where they can be identified and recruited to participate. In Texas, lists of permitted construction sites are available through a public information request. Lists were requested monthly, and sites served as the "venue" for sampling frame. The city was broken into 12 regions, and sites were sampled, by simple random sampling, from one region per month, every month, for a year. Permitted sites were randomly selected and assigned hour-long windows for field staff to visit those sites. Although permits are required for construction, they are not always pulled.

To capture nonpermitted construction sites, the study team also canvassed blocks surrounding the selected sites to identify other visible construction sites that were not sampled. The construction sites that were visited for the TLS sample were mostly residential construction (e.g., townhomes, single-family homes), with some commercial buildings (e.g., large pieces of land where a new building was being constructed, high-rises). Field interviewers (FIs) approached workers at each site, screened them for eligibility, and administered the web-based survey on a tablet. Administration occurred in three ways, depending on each participant's preference: (1) the participant self-administered the survey on the tablet, (2) the FI administered the survey by reading the questions and response options to the participant verbatim, or (3) the FI provided the worker with information to complete the survey later on a personal device. The survey took approximately 10-20 minutes to complete. Participants were provided with their choice of a physical or electronic $50 gift card for their participation.

LTS

LTS was used to supplement the TLS sample. LTS is a network sampling approach that relies on study participants to recruit their peers to participate in the study. LTS is similar to respondent-driven sampling (RDS), which has been used successfully in studies to estimate the prevalence of sex trafficking (e.g., Dank et al., 2019; Jordan et al., 2018) and labor trafficking of undocumented immigrants (Zhang et al., 2014). LTS is a promising method to use among construction workers because they often work in groups and have regular contact with other construction workers who may be eligible to participate in the study. The peer recruitment process starts by selecting a set of initial participants (i.e., seeds) who complete the survey and then invite their eligible peers to participate. Seeds included members of the TLS sample and additional respondents recruited at locations where day laborers congregate (e.g., Home Depot, gas stations).

Much like the TLS sample, the project team monitored data collection closely to adjust for any needed changes. Data collection began by inviting every third worker to be the seed. By the end of data collection, every second worker was the seed. Like the TLS sample, all seeds were provided the option of a physical or electronic $50 gift card for completing the survey. The seeds were then allowed to invite up to three eligible peers to participate, each of whom could also invite up to three peers. At the end of the survey, participants were asked whether they would be willing to provide contact information (phone or email) for friends or family they knew that also worked in construction in the past 2 years. If they agreed, participants entered the information for up to three peers. Participants were provided a $25 electronic gift card for each referral (up to a maximum of $75 for three peers) who completed the survey. Because the referral process occurred without the FI present, a physical gift card was not an option, and participants were required to enter their email address to receive the incentive for completing the survey.

Time Method View help for Time Method

Universe View help for Universe

Adult persons who worked in the construction industry within the past 2 years.

Unit(s) of Observation View help for Unit(s) of Observation

Data Type(s) View help for Data Type(s)

Mode of Data Collection View help for Mode of Data Collection

Description of Variables View help for Description of Variables

The data includes variables on the types of work participants perform in the construction industry, workplace conditions, recruitment experiences, and treatment by employers, including instances of feeling pressured to accept jobs. Demographic variables include race, gender, age, primary language, and marital status.

Original Release Date View help for Original Release Date

2025-09-29

Version History View help for Version History

Weight View help for Weight

TLS

Sampling weights were calculated as the inverse product of two-stage selection probabilities using the Horvitz-Thompson estimator for TLS samples following the approach of Leon et al. (2015). Leon et al. describe a three-stage TLS estimator wherein the first stage was the selection of locations. In the current study, all zones were visited (i.e., excluding subsample zones) and thus have equal probabilities, reducing the research team's implementation of the Horvitz-Thompson TLS estimator to two stages: The selection of sites within geographies and the selection of workers within each site. The probability of selecting a site within a geography was calculated as the number of selected sites divided by the number of permitted sites available to visit. The probability of worker selection was 1/n where every nth worker was selected. When a site had no workers, interviewers canvased the area for other sites, which generated a substantial number of additional interviews as reported in the results. An analog TLS sampling weight was developed for these individuals to allow for estimates using a combination of the randomly selected TLS sample and the convenience samples (or unselected sites) obtained when canvassing near selected sites.

LTS

Sampling weights for the network sample were derived through a statistical matching procedure, which was tested on the TLS sample to ensure the reliability of the approach. Given the lack of direct selection probabilities for the network sample, pseudo-probabilities were imputed using a beta regression model. The model incorporated key covariates, including the zone of the corresponding site, gender, disaster site worker status, and ratio of workers present to those interviewed. The imputed probabilities were then used to generate sampling weights for the network respondents. As with the TLS sample, some of the imputed weights were extreme and were therefore trimmed to mitigate their influence on the estimates. The trimming procedure involved capping weights at five times the mean and one-fifth of the mean, following the guidance of Battaglia et al. (2004). This adjustment helped stabilize the estimates and reduce the potential impact of outliers. The final trimmed weights were used in the analysis to calculate weighted estimates for the network sample.