Methods note: quality control
How do you catch bad data while fieldwork is still running?
Nigeria's survey data carry well-documented problems with ages and dates, and where interviewers differ, a few of them often account for most of the problems. The checks that find them in time to fix them.
Published October 2026. Sources checked 11 October 2026.
In short
- Age heaping in Nigeria's DHS surveys from 2003 to 2013 was among the highest of any DHS survey of 2000–2015, and a DHS review places Nigeria among the countries with the poorest age and date data.
- Training can over-correct. After earlier surveys moved young women out of the eligible age range, the 2003 and 2013 surveys moved them the other way. Checks have to look in both directions.
- Where interviewers differ, the problems can often be traced to a handful of them. In South Africa's national panel study, checks found during fieldwork that about 7% of the sample had been fabricated, and the interviews were redone.
Ages and dates in Nigerian surveys
The DHS Program's methodological reviews document the problem in detail. Myers' Blended Index estimates the percentage of reported ages that would have to move to remove the preference for particular final digits. For Nigeria's household populations it was 14.3 in 2003, 14.2 in 2008 and 13.5 in 2013, and among DHS surveys of 2000–2015 only India's 2005–06 survey scored higher than Nigeria's 2003 and 2008 surveys1. Combining incompleteness, heaping and displacement, the review places Nigeria among 13 countries in the poorest fifth for the quality of ages and dates1.
Ages near an eligibility boundary are also shifted across it. Nigeria's 1990 survey showed the highest level of any DHS survey of 1985–2003 of young women recorded as younger than 15, outside the eligible range2. The 2003 and 2013 surveys then showed large shifts the other way, which the DHS review attributes to 'an over-correction' after efforts to prevent the earlier problem1. Heaping also shows at the upper boundary: the 2008 household listing records 407 women aged 49 and 914 aged 50, though men show an even sharper jump (286 and 1,157), and the 2017 review did not flag Nigeria for transfers across age 503,1.
Birth dates show a similar pattern. In the 2008 survey, births recorded for odd years from 1999 to 2003 far outnumbered those for the even years beside them, with calendar-year ratios for 1999–2004 swinging between 73.5 and 139.5, where 100 means no difference3. A DHS review of fertility estimates classes Nigeria's data as 'poor'4.
Problems often trace to a few interviewers
A DHS study of 25 data-quality indicators in 15 surveys, Nigeria's 2015 malaria indicator survey among them, found that where interviewers differ a lot, 'the bulk of that variation can often be traced to a handful of interviewers'. Older and better-educated interviewers generally produced fewer problems. The study warns that interviewers usually work in one or two regions, so their effects cannot be cleanly separated from those of the places they work5.
Fabrication is the extreme case. In South Africa's National Income Dynamics Study, fabrication affected about 7% of the sample; it 'was detected while fieldwork was still on-going, and the relevant interviews were reconducted'. The fabricated data would not have changed cross-sectional estimates much, but would have led to 'substantially different conclusions' in panel analysis, and the authors put the benefit-cost ratio of the checks at 'at least 24'6.
The checks, and what the numbers mean
National surveys already check data during fieldwork. The DHS produces field check tables on response rates, age distributions and missing values while teams are still in the field, and discusses problems with them7. Nigeria's 2018 DHS generated these tables weekly8, and the 2024 DHS reviewed them in weekly meetings, with tablet software that accepted only valid responses and checked ranges automatically9. Nigeria's GHS-Panel monitored its data daily through the World Bank's Survey Solutions software10.
Back-checks, short re-interviews by someone else, are the standard defence against fabrication. J-PAL recommends back-checking 'at least 10% of respondents, as a best practice', with every interviewer back-checked at least once11. The World Bank's DIME guidance says to aim for 10–20%, with 20% of back-checks in the first two weeks, by a team 'independent from the rest of the survey staff'12. These are rules of thumb; neither cites evidence for the numbers.
Tablets add evidence of their own: a log of every answer and edit, with its time13; optional audio recording of all or part of an interview14, which we use only with ethics approval and the respondent's consent; and GPS readings. GPS shows where a device was, not that an interview took place. In northern Nigeria in 2010, GPS tracks of polio vaccination teams 'revealed that teams commonly missed swaths of contiguous households'15.

Our practice
These are the rules every Probatio Africa study follows. Read them as commitments you can hold us to.
- We run checks on every day's data, by interviewer as well as by team: duplicates, interview lengths, skipped sections, outliers, and ages on both sides of each eligibility boundary.
- We back-check at least 10% of interviews, every interviewer at least once and more in the first two weeks, with back-checkers independent of the team, comparing answers that should not change.
- We do not train on a single indicator. We watch for shifts in both directions across the 14/15 and 49/50 boundaries and in years of birth.
- When checks show a problem, we pause the interviewer or the team, retrain or replace, and re-interview where needed. Every check and its result goes into the study file.
- We record interviews only where the ethics committee approves it and the respondent agrees.
- Every report states the quality-control steps we took, as the AAPOR Code requires16, and what they found.
What the evidence cannot settle
- Whether age and date reporting improved in Nigeria's 2018 and 2024 DHS. The published reviews stop at 2015, and we have not yet analysed the newer data.
- The right back-check rate. The 10–20% range is professional guidance; neither source we cite gives evidence for it.
- How effective audio checks are in African surveys. The evidence we found is from the United States.
Sources
- 1.T. Pullum and S. Staveteig, DHS Methodological Reports No. 19, ICF (August 2017). An assessment of the quality and consistency of age and date reporting in DHS surveys, 2000–2015.https://dhsprogram.com/pubs/pdf/MR19/MR19.pdf
- 2.T. W. Pullum, DHS Methodological Reports No. 5, Macro International (December 2006). An assessment of age and date reporting in the DHS surveys, 1985–2003.https://www.dhsprogram.com/pubs/pdf/MR5/MR5.pdf
- 3.National Population Commission and ICF Macro (2009). Nigeria Demographic and Health Survey 2008, Appendix D: data quality tables. Tables D.1 and D.4.https://nigerianstat.gov.ng/nada/index.php/catalog/27/download/90
- 4.B. Schoumaker, DHS Methodological Reports No. 12, ICF International (September 2014). Quality and consistency of DHS fertility estimates, 1990 to 2012.https://www.dhsprogram.com/pubs/pdf/MR12/MR12.pdf
- 5.T. W. Pullum, C. Juan, N. Khan and S. Staveteig, DHS Methodological Reports No. 24, ICF (September 2018). The effect of interviewer characteristics on data quality in DHS surveys.https://www.dhsprogram.com/pubs/pdf/MR24/MR24.pdf
- 6.A. Finn and V. Ranchhod, World Bank Economic Review 31(1): 129–157 (2017). Genuine fakes: the prevalence and implications of data fabrication in a large South African survey.https://doi.org/10.1093/wber/lhv054
- 7.ICF International (December 2012). Survey Organization Manual for Demographic and Health Surveys.https://dhsprogram.com/pubs/pdf/DHSM10/DHS6_Survey_Org_Manual_7Dec2012_DHSM10.pdf
- 8.National Population Commission and ICF, via the World Bank Microdata Library (Accessed 11 October 2026). Nigeria Demographic and Health Survey 2018: study description.https://microdata.worldbank.org/catalog/3540/study-description
- 9.Federal Ministry of Health and Social Welfare, National Population Commission and ICF (September 2024). Nigeria Demographic and Health Survey 2023–24: Key Indicators Report. Section 2 describes the sample, fieldwork and the clusters not reached.https://dhsprogram.com/pubs/pdf/PR157/PR157.pdf
- 10.National Bureau of Statistics and World Bank (Accessed 11 October 2026). General Household Survey, Panel 2023–2024, Wave 5: study description.https://microdata.worldbank.org/catalog/6410/study-description
- 11.M. Gibson, J-PAL Research Resources (Updated March 2021). Data quality checks.https://www.povertyactionlab.org/resource/data-quality-checks
- 12.World Bank DIME Wiki (16 August 2023). Back checks.https://dimewiki.worldbank.org/Back_Checks
- 13.World Bank, Survey Solutions documentation (Accessed 11 October 2026). Paradata file format.https://docs.mysurvey.solutions/headquarters/export/paradata_file_format/
- 14.World Bank, Survey Solutions documentation (Accessed 11 October 2026). Audio audit.https://docs.mysurvey.solutions/headquarters/svymanage/audio-audit/
- 15.V. M. Gammino and others, Journal of Infectious Diseases 210(Suppl 1): S98–S101 (2014). Using geographic information systems to track polio vaccination team performance: pilot project report.https://doi.org/10.1093/infdis/jit285
- 16.American Association for Public Opinion Research (As amended June 2026). The Code of Professional Ethics and Practices.https://aapor.org/wp-content/uploads/2026/06/AAPOR-2026-Code-of-Ethics-Final-.pdf
How to cite this note
Probatio Africa (2026). How do you catch bad data while fieldwork is still running? Methods note. Sources checked 11 October 2026. https://probatioafrica.com/evidence/methods/checking-data-during-fieldwork.
Our text is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0): you may copy, share and adapt it, including in paid work, if you credit Probatio Africa and link to this page. Quotations from laws, standards and other sources keep their owners' terms; cite the original when you reuse one. If you find an error, write to enquiries@probatioafrica.com. We correct errors quickly and record each correction, with its date, at the foot of the note.
Related notes
Every note follows the same format and is checked against its sources.
Questionnaires
How do you ask the same question in Hausa, Yorùbá, Igbo, Pidgin and English?
Translation is where comparability is easily lost and seldom documented. What Nigeria's national surveys do, where the standards disagree, and the procedure we follow.
Read the noteSurvey modes
When can a phone survey stand in for a household visit?
Phone surveys are fast and cheap, and in Nigeria they can reach most households. They miss particular people, and weighting narrows the gap without closing it.
Read the note