Living Standards Measurement Study

Select a EDS Sub navigation page selecting option, leaving this page
Follow Us

SUBJECTIVE WELFARE

The World Bank
In recent years there has been increasing recognition that poverty is a multifaceted and multidimensional problem, and that traditional one-dimensional measures, such as per-capita consumption, do not sufficiently capture the complexity of the subject. In response, subjective welfare measures have been increasingly employed to provide additional measures of well-being. This component of the research program has two aims. First, it will review past questions and the experience with those questions from the literature on subjective welfare in both developed and developing countries. Based on these findings, it will propose a preliminary core set of subjective welfare questions. Second, the project will aim to make a significant advance in our knowledge about how best to collect such questions in developing countries and how best to validate and adjust the data for inherent comparability problems.

Anchoring Vignettes

One-dimensional measures of poverty, such as per-capita consumption or income, do not sufficiently capture the complexity and multidimensionality of poverty. Subjective welfare measures, in which respondents self-report their household’s welfare level, offer alternative measure of living standards. This component of the research program focuses on improving subjective measures of well-being. The project aims to improve these measures, including how to validate and adjust the data for inherent comparability problems.

While subjective questions have been widely used in fields such as psychology and sociology for many years, some question their applicability to the measurement of well-being. The main critique of their validity is that unobservable characteristics of survey respondents may influence the scales people use to assess if they are poor or rich. Anchoring vignettes may potentially correct for the resulting bias in subjective reports of welfare. This approach has been adapted from Harvard University political scientist Gary King and co-authors; in their work, these vignettes are applied to self-reports of health status and political efficacy. Here we extend this to subjective welfare reports. Anchoring vignettes provide common points of comparability across heterogeneous groups, and can be used to rescale subjective questions and reduce bias from unobservable characteristics. To date, in this LSMS research program, anchoring vignette have been piloted and fielded in household surveys in Tanzania, Tajikistan and Guatemala. The English text of these vignettes is available here.

Experiments

Anchoring vignettes in Tanzania

We piloted and fielded anchoring vignette and subjective welfare questions in the Kagera Subjective Welfare Survey, a household survey conducted in November-December 2007 in the Ngara district of Tanzania. As the first set of vignette we developed, significant time was devoted in June 2007 with respect to the topics and text of the vignettes. The vignettes of the vignettes was developed in consultation with local consultants. Topics were chosen to measure locally relevant information, such as food security, quality of medical care, quality of education, and poverty. The total sample size was 450 households. Data was collected using Ultra Mobile Personal Computers.

More Survey Methods

Of the 1.4 billion people living in extreme poverty, the vast majority resides in rural areas, relying on smallholder agriculture as a source of income and livelihood. The FAO estimates that Africa is home to 33 million small farms, holding less than two hectares and representing 80 percent of all farms.

Farming practices are typically very labor intensive and the majority of the labor is provided by household members. Agricultural household labor is therefore a key household asset and its accurate measurement is important. The estimation of labor inputs on smallholder farms is complex and vulnerable to misreporting.

Smallholder farms typically employ mostly family labor; thus, there is no wage income in which to anchor recall estimates. Written records are rarely kept and the respondent must rely on recall to report on past events. To arrive at the total amount of labor allocated by a household to farming, the household must accurately report the plots under cultivation, the specific household members that worked on each plot, the activities performed, and their timing and duration. Farming is a seasonal activity and work patterns are irregular during the season.

Reporting “typical” or “average” time farming after the completion of the season requires remembering distant events and making complicated mental calculations. Alternatively, reporting hours worked in the last 7 days at any single point during the agricultural season will not necessarily be indicative of total labor during the season if labor inputs vary considerably across weeks during the season.

Experiments

Tanzania Farm Labor Experiment to study the accuracy of farm labor data in household surveys, we conducted a survey experiment during the main agricultural season (roughly January-June 2014) in the Mara district of Tanzania. A random sample of 854 households from 18 communities was randomly assigned to one of the following alternative survey designs:

  1. Weekly Visit (benchmark): weekly face-to-face surveys for the duration of the season.
  2. Weekly Phone (alternative): weekly phone surveys for the duration of the season
  3. Recall Modules NPS (business-as-usual): single face-to-face survey at the end of the agricultural season.

Two commonly used designs were tested. We established the magnitude of bias by comparing the Weekly Phone and Recall groups to the Weekly Visit design.

Malawi Farm Labor Experiment to study the accuracy of farm labor data in household surveys. The experiment was conducted in the Zomba and Ntcheu districts both of which are in southern Malawi. The study covered a total of 20 enumeration areas (EAs) – 10 EAs in Zomba district and 10 EAs in Ntcheu districts. These EAs were randomly selected from all rural EAs in each district. The experiment started with three treatment arms based on data collection approaches, with a fourth arm added at endline survey phase. First, the Control Group 1 (C1) in which the standard agricultural labor module, with labor reported in the aggregate by recall for the entire season (endline survey) was used as a data collection approach with both baseline and endline surveys administered through face-to-face interviews. Secondly, the Treatment Group 1 (T1) in which data collection was based on weekly phone surveys for labor module for the duration of the rainy season. Thirdly, the Treatment Group 2 (T2) in which an intensive interview labor module (Time-use survey) during the duration of the main season was administered every week. Fourthly, Control Group 2 (C2) in which intensive interview of labour module was administered at the end of the agricultural season only (only included in the endline survey).

The data collection was divided into three phases:

1. Baseline survey phase  (August 2016 - October 2016)

2. Resident enumerator survey (weekly data collection) phase (November 2016 - May 2017)

3. Endline survey phase (June 2017 - August 2017)

Resources

  • Not your average job: Measuring farm labor in Tanzania (Paper)

  • Not your average job: Measuring farm labor in Tanzania (Brief)

  • Measuring family farming is tricky business (Blog)