2.2 Background
2.2.2 DHS Interviewers: Contracts and Incentives
This section describes the process of hiring DHS interviewers and conducting surveys, examining the sources of workplace incentives and describing the tasks the workers complete.
The DHS program has facilitated the administration of over 300 nationally and regionally representative surveys of reproductive and health behaviors in developing countries since the 1980s. The surveys are administered by local implementing agencies with technical support from the DHS program, which is funded by the U.S. Agency for International Development.
The implementing agencies are most often governmental organizations such as National Statistical Offices or the Ministry of Health, but can also be non-governmental or private.
The DHS program sets standard guidelines for the implementation of the surveys: manuals are publicly available for interviewers and supervisors and editors for the processes of training field staff and sampling, among other topics. The questionnaires used in the DHS are also standard within phases of the survey, although optional modules (e.g. domestic violence, HIV, and anemia) are implemented in some places but not others. These standardized procedures are put into place to ensure comparability of the data across countries. This means that interviewers’ tasks are standardized across contexts, though their employers vary.
7In Section 2.5 I show direct evidence of a significant interaction effect between temperature and humidity in determining worker productivity.
The focus of the surveys is on the collection of accurate reproductive histories and information on the health status of women and children in developing countries. Therefore, the majority of respondents are prime-aged women (between the ages of 15 and 49), although in many survey rounds some males are interviewed in the households as well. In the majority of the more recent DHS surveys, DHS teams have also been collecting GPS coordinates for each survey cluster (e.g. a village in a rural area or a city block in an urban area) in the sample.8 Most of these are accurate to within 15 meters.9
Recruitment of fieldworkers is conducted locally by implementing agencies, but the practices for recruitment and required qualifications are standard across contexts. In most cases, interviewers are temporary employees of the survey implementing organization for the duration of the survey, although sometimes the implementing agencies use their own permanent staff.10 Interviewers must be available to work full time for the duration of the survey, including nights and weekends, and they must have sufficient physical fitness to walk long distances and carry questionnaires as required. There is a strong preference for interviewers to interview respondents of their own gender, and they must speak at least one of the languages used for training (and thus prominent in that local context). Interviewers are recruited within a region of a country as much as possible so that the interviewer does not seem too foreign to the respondent, and there is a preference for candidates with at least a secondary education. This means that in most cases, interviewers are a more highly-educated population than the average residents of their countries.11
8Clusters, or enumeration areas, are generally chosen with probability proportional to population size from a recent census.
91% of rural clusters are displaced between 0 and 10 kilometers to protect the privacy of the respondents.
10This is not observable in this data sample.
11Information on interviewer characteristics beyond the unique identifier is not available in most rounds of
The contracts in each survey round are determined by the survey implementing agency, which has freedom in designing pay structure. However, conversations with technical staff at ICF, which implements the DHS, indicate that the standard practice in the DHS is to pay interviewers a fixed amount per day plus a per diem for food/lodging/etc. Anecdo-tally, piece-rate wages are problematic in this context because some interviews naturally take longer than others, so piece rates would introduce significant risk into interviewers’ wages, and furthermore, piece rates incentivize quantity of production over quality. The fact that interviewers are not paid piece rate, however, means that their incentive to perform well comes from any risk of termination, or from wanting to maintain a good reputation (many interviewers have previous experience and will go on to work for other surveys). There is no direct information on how many workers are fired during the average survey, but the imple-menting organizations are recommended to hire 10 percent more interviewers than needed to serve as reserve interviewers to fill gaps in case of separations.12
the DHS. In a few recent surveys, the DHS has included a questionnaire on interviewer characteristics.The surveys are: Afghanistan, 2015-2016; Armenia, 2015-2016; Nepal, 2016; Zimbabwe, 2015. These surveys took place after 2010, so they cannot be linked with the weather data. However, I summarize the information on interviewer education and age in Figure B9. The data reveal that levels of education actually vary by context in the DHS; interviewers in Africa have few years of education while interviewers in Afghanistan and Nepal are quite well educated. The interviewers are very young on average, and the data also reveal that almost half had worked on a prior DHS survey and nearly 75 percent had worked on another survey of some kind.
12In Table B1, I examine whether interviewers are more likely to “leave” the survey round before the rest of their team if they have worse average measures of productivity. I am unable to observe whether these separations occur because the interviewer, quit, was fired, or was simply finished with his or her interviews, but I document a significant correlation between interviewer productivity and having a final interviewing date before the rest of the interviewer’s team.
2.2.2.1 Interviewer Tasks and Performance Evaluation
Data are collected by interviewing teams usually comprised of a supervisor, a female field editor, several female interviewers, and one to two male interviewers if men will be included in the survey. Supervisors and field editors often share responsibilities, but super-visors in general are responsible for organizing fieldwork (arranging transportation, lodging, etc.), delegating interviews to interviewers, and conducting spot check re-interviews. The focus of field editors is on data quality: field editors observe at least one interview per day and edit completed questionnaires while in the field, sending interviewers back to correct problems if necessary.13 Each interviewing team generally is assigned its own car.
Data collection in each household begins with a household survey, in which basic information on each member of the household is recorded along with descriptive information on the household such as building materials and water sources. In selected households, all women aged 15-49 are then eligible to be interviewed using the individual questionnaire. In many survey rounds, men are also interviewed, usually in every second or third household.
Reproductive history is gathered from both men and women, but only women answer de-tailed questions about child health and other outcomes. In interviews with women, dede-tailed questions are asked about all children under an age cutoff, usually five. The DHS also in-cludes anthropometric measurements of children under five and women, such as height and weight and often country-specific biomarkers such as tests for anemia, malaria, or HIV. In cases where either an individual or an entire household is not available for interview on the first visit, DHS rules stipulate that interviewers must make at least 3 visits at separate
13It is not observable in the DHS data which interviews were observed by a field editor.
times/days to make every effort to interview eligible individuals. In the data, the DHS iden-tifies the interviews that required more than one attempt, although there is no information on when the unsuccessful attempts took place. As I discuss in Section 2.6, any effect of temperature on the probability of nonresponse could affect the main results on quantity and quality of data production. I investigate this empirically below.
The schedules of DHS interviewers are set at the beginning of each day by the team’s field supervisor. Guidelines indicate that the supervisors give assignments in the form of lists of households that each interviewer is meant to visit, but that assignments are often updated throughout the day, as some households take longer or shorter amounts of time than expected to interview. Performance is evaluated continuously throughout the survey round using an “interviewer progress sheet,” shown in Figure 2.1. The supervisors keep one progress sheet per interviewer throughout the fieldwork process and update the sheet after each cluster to ensure that each member of their team is keeping up with the assigned workloads. I therefore examine number of completed interviews as a very highly-monitored measure of interviewer productivity.
The DHS has several other mechanisms in place for monitoring interviewer perfor-mance. As mentioned previously, field editors regularly observe individual interviews (they are responsible for one observation session per day). Supervisors conduct one reinterview per cluster to ensure that interviewers are not engaging in common practices meant to lighten workloads, such as interviewing smaller households that were not selected or classifying el-igible individuals as inelel-igible for individual interviews by misreporting their ages. Finally, field editors check questionnaires in the field, providing an additional check on the quality of data produced by each interviewer. The data quality problems observed in the final data
are flagged in secondary processing after fieldwork is complete; they are issues that were not caught during fieldwork. These flags for incomplete, inconsistent, or missing data comprise the outcome variables used to study production quality below.