The Survey of Household Welfare and Labour in Tanzania (SHWALITA) is a 4,000-household survey, funded by the World Bank and implemented by EDI Global in 2007. It comprises three separate experiments:
(i) Consumption experiment: This experiment involves eight alternative consumption questionnaires distributed randomly among 4,000 households. The questionnaires differ in method (diaries or recall modules), length of recall period, and number of items.
(ii) Labour module experiments: These experiments examine the collection of labour statistics using different approaches. Two modules, a long one and a short one, were administered to either the individual or a proxy respondent in the household. Both self-reporting and proxy respondents were randomly sampled from the household roster.
(iii) Subjective welfare experiments: These experiments employed an innovative approach to enhance comparability of subjective welfare questions. The technique, developed in political sciences by Gary King, involves the respondent to provide scaled answers on qualitative questions (on a scale of 1 to 5, how do you feel about….). In order to ‘anchor’ the response the respondent is given a ‘vignette’ a short, but powerful story about a fictitious person and is then asked to place this person on the same scale. The placing of the vignette on the same scale allows answers to become more comparable across households, communities and countries.
The questionnaire and data are available on the World Bank’s website, here. Publications using SHWALITE are available here.
Rationale
The rationale behind the CONSUMPTION EXPERIMENTS comes from the observation that there are large and growing gaps between micro and macro estimates of household consumption. These discrepancies have profound implications for measuring global progress in poverty reduction and the effect of economic growth on that process. Currently it is difficult to reconcile these differences due to the wide variation in methods used to measure household consumption. While macro measures are broadly consistent around the world, under the SNA framework, micro measures of household consumption have no such standardization. Household expenditure surveys vary widely across many dimensions, including: the method of data capture (diary versus recall), the level of respondent (individual versus household), the length of the reference period for which consumption is reported (varying from 3 days, to one week, to one year) and the degree of commodity detail in recall surveys (varying from less than 20 to over 400 items). These variations occur both across countries and also over time as statistical offices alter survey design, with little understanding of the implications of such changes for spatially and temporally consistent measurement of household consumption and poverty. This variation hampers both cross-country studies of poverty and well-being measures as well as measuring poverty trends within country. This experiment implements alternative methods to measure household consumption.
The researchers developed eight alternative consumption questionnaires which were randomly distributed across 4,000 households. These eight designs vary by method (3 diaries and 5 recall modules), length of reference period in recall modules, and the number of items in the recall modules. In addition to assessing how the alternative methods affect consumption calculations and household rankings, the evaluation will include a comparison of costs across numerous dimensions: length of field work (in part based on length of interview which will be recorded), coding and data entry inter alia. The study also assesses the sensitivity (robustness) of poverty line calculations where the food poverty is based on calorie assignment of food groups in turn affected by level of disaggregation of food items.
The LABOUR EXPERIMENTS assess the effect of different ways of collecting labour statistics. It uses two different modules, a long module and a short module, and administers each to either the person him/herself or to someone else in the household answering on their behalf (a proxy respondent). Both proxy respondents and self-reporting respondents are sampled randomly from the roster of household members.
The SUBJECTIVE WELFARE EXPERIMENTS use an innovative approach to enhance comparability of subjective welfare questions. The technique, developed in political sciences by Gary King, involves the respondent to provide scaled answers on qualitative questions (on a scale of 1 to 5, how do you feel about….). In order to ‘anchor’ the response the respondent is given a ‘vignette’ a short, but powerful story about a fictitious person and is then asked to place this person on the same scale. The placing of the vignette on the same scale allows answers to become more comparable across households, communities and countries. Data were captured electronically through CWEST.
Sampling & Module Assignment
The 7 districts covered in this project were previously surveyed through EDI’s CWIQ project (see tab at the top of this page for more detail), in which a sample of households was drawn to be representative at district level. Data from the 2002 Census was used to put together a list of all villages in the district. In the first stage of the sampling process villages were chosen proportional to their population size. In a second stage the sub-village (kitongoji) was chosen within the village through simple random sampling. In the selected sub-village, or cluster all households were listed. Shwalita makes use of CWIQ’s sampling frame to randomly select 24 clusters out of the 30 CWIQ clusters and draw its random sample of households from the CWIQ listing forms. The following table shows the selected districts and is sorted in the order in which they will be visited.
District | region | urban/rural | adult literacy rate according to CWIQ | Available CWIQ documents |
Bukoba Rural | Kagera | rural | 81% | report – brief ENG – brief SWA |
Karagwe | Kagera | rural | 67% | report – brief ENG – brief SWA |
Bukombe | Shinyanga | rural | 62% | report – brief ENG – brief SWA |
Bariadi | Shinyanga | rural | 54% | report |
Rufiji | Pwani | rural | 61% | report – brief ENG – brief SWA |
Temeke | Dar es Salaam | urban | 90% | report – brief ENG – brief SWA |
Dodoma Urban | Dodoma | urban | 75% | report |
The following 8 modules are randomly assigned to 3 households within each cluster:
Consumption Recall and Labour Modules:
module No. | type of labour module | recall length in consumption module | type of item list in consumption module | total sample size(24 clusters in each of 7 districts) | downloads questionnaires |
1 | short labour module with reporting by proxy respondent | 14 days | long item list | 504 obs.(1/3 without labour module) | ENGLISH – SWAHILI |
2 | short labour module with members self-reporting | 7 days | long item list | 504 obs.(1/3 without labour module) | ENGLISH – SWAHILI |
3 | long labour module with reporting by proxy respondent | 7 days | short subset of long list | 504 obs.(1/3 without labour module) | ENGLISH – SWAHILI |
4 | long labour module with members self-reporting | 7 days | short collapsed list (aggregation of items from long list) | 504 obs.(1/3 without labour module) | ENGLISH – SWAHILI |
5 | NONE | 1 month |
long item list504 obs.ENGLISH – SWAHILI
Diaries:
Module No. | level at which administered | diary period | frequency of visits by interviewer | frequency of visits by locally recruited assistant | total sample size(24 clusters in each of 7 districts) | downloads |
6 | individual | 14 days | frequent visits:all individuals on days 1-3-5-8-10-12-15 | every day | 504 | ENGLISH – SWAHILI |
7 | household | 14 days | frequent visits:all households on days 1-3-5-8-10-12-15 | every day | 504 | ENGLISH – SWAHILI |
8 | household | 14 days | infrequent visits:Literate households: days 1-8-15.Illiterate households days 1-3-5-8-10-12-15 | no visits | 504 | ENGLISH – SWAHILI |
Download diary household questionnaire (administered during first and last vist): ENGLISH – SWAHILI
Finally, the subjective welfare questionnaire will be administered to 576 households (4 households in each of 24 clusters in each of 6 districts) and will be downloadable from this site soon.
The survey teams will visit 168 communities. In each community the nearby shops, stalls and markets will be visited to collect local price data (download price questionnaire). Additionally, a structured community questionnaire will be administered to key informants in each community (download English – download Swahili). The community questionnaire contains a price opinion section as an alternative way to collect prices. For a good discussion on various price collection mechanisms in surveys see Gibson and Rozelle’s WBER article.
Timing
EDI began piloting questionnaires and training interviewers from June 2007 onwards. Fieldwork started beginning of September 2007 and is expected to last till end of June 2008. In order to keep tight control implementation, the fieldwork is conducted by a relatively small number of 12 interviewers and spread over a longer time period. Such a set-up avoids the typical co-ordination problems faced by larger-scale fast-moving set-ups and allows for maximum control from the project management and co-ordination team. Ultimately it seems like a necessary condition to achieve an acceptable level of non-sampling error.
This assignment is being executed by the following members of staff at EDI:
Project Direction: Joachim De Weerdt
Management and Co-ordination: Respichius Mitti and Abida Nungu
Field Supervision: George Musikula, Davis Matovu, Josephine Rugomora and Pius Sosthenes
Enumeration: Abbanove Gabba, Aissa Issa, Faustine Misinde, Felix Kapinga, Geofrey Bakari, Honoratha Wyclife, Jamary Idrisa, Jesca Nkonjelwa, Kamugisha Robert, Makarius Kiyonga, Resty Simon, Hildephonce Muhashani
Data Entry Co-ordination: Thadeus Rweyemamu
Data Entry Operation: George Gabriel, Justina Katoke, Amina Suedi, Frida George
Publications resulting from this project:
Gazeaud, Jules. 2017. Are PMT Performances Vulnerable to Measurement Errors in Consumption? Evidence from a Survey Experiment in Tanzania. Mimeo. CERDI, University of Auvergne. Download paper
Amaye, Hannah. 2017. Urbanization and the Two Tails of Malnutrition in Tanzania. LICOS DIscussion Paper. Download paper
Friedman, Jed, Kathleen Beegle, Joachim De Weerdt and John Gibson. Forthcoming. Decomposing Response Error in Food Consumption Measurement: implications for survey design from a randomized survey experiment in Tanzania”. Food Policy. World Bank Policy Research Working Paper 7505.
Dillon, Brian, Joachim De Weerdt and Ted O’Donoghue. 2016. “Paying More for Less: Why Don’t Households in Tanzania Take Advantage of Bulk Discounts?” Mimeo
Ravallion, Martin, Kristen Himelein, and Kathleen Beegle. 2016. “Can Subjective Questions on Economic Welfare Be Trusted? Evidence for Three Developing Countries.” Economic Development and Cultural Change. Economic Development and Cultural Change 64(4): 697–726
Ravallion, Martin, Kristen Himelein, and Kathleen Beegle. 2016. “Can Subjective Questions on Economic Welfare Be Trusted? Evidence for Three Developing Countries.” Economic Development and Cultural Change. World Bank Policy Research Working Paper 6726.
De Weerdt, Joachim, Kathleen Beegle, Jed Friedman and John Gibson. 2016. The Challenge of Measuring Hunger through Survey. Economic Development and Cultural Change, 64(4): 727–758. download pdf
Friedman, Jed, Kathleen Beegle, Joachim De Weerdt and John Gibson. 2015. Decomposing Response Error in Food Consumption Measurement: implications for survey design from a randomized survey experiment in Tanzania”. Mimeo.
Gibson, John, Kathleen Beegle, Joachim De Weerdt and Jed Friedman. 2015. What Does Variation in Household Survey Methods Reveal About the Nature of Measurement Errors in Consumption Estimates? Oxford Bulletin of Economics and Statistics 77(3): 466-474.
Beegle, Kathleen, Joachim De Weerdt, Jed Friedman and John Gibson. 2012. Methods of Household Consumption Measurement through Surveys: Experimental Results from Tanzania. Journal of Development Economics 98:3-18.
Bardasi, Elena, Kathleen Beegle, Andrew Dillon and Pieter Serneels. 2011. Do Labor Statistics Depend on How and to Whom the Questions are Asked? Results from a Survey Experiment in Tanzania. World Bank Economic Review 25(3): 418 – 447
Dillon, Andrew, Elena Bardasi, Kathleen Beegle and Pieter Serneels. 2012. Explaining Variation in Child Labor Statistics. Journal of Development Economics, 98 (1): 136-147.