Demographic and Health Survey Analysis
- Introduced by
- USAID / The DHS Program (ICF)
- Year
- 1984
- Type
- Nationally representative population and health survey
- Family
- Health and demographic survey
What does it do?
The Demographic and Health Surveys (DHS) are nationally representative household surveys that provide standardised, internationally comparable data on population, health, and nutrition in low- and middle-income countries.
Funded primarily by USAID and implemented through The DHS Program, they use model questionnaires, a complex multi-stage sample design, and a standardised wealth index to produce indicators of fertility, child and maternal mortality, family planning, child nutrition, and disease prevalence that drive health policy and programme monitoring worldwide.
The intuition behind the method
The power of the DHS comes from doing the same thing the same way everywhere. By administering near-identical model questionnaires through a carefully documented sampling and weighting procedure, the programme makes a stunting rate in Malawi directly comparable to one in Bangladesh, and a 2015 figure comparable to one from 1995. Because money income is hard to measure, the survey instead builds a wealth index from durable goods and housing characteristics to rank households, and it relies on women's birth and pregnancy histories to reconstruct fertility and child mortality without needing a vital registration system.
History
The DHS Program began in 1984, funded by the U.S. Agency for International Development, building on the earlier World Fertility Survey and Contraceptive Prevalence Surveys. Successive phases (DHS-I through DHS-8) standardised and expanded the model questionnaires, adding HIV and other biomarker testing, GPS cluster coordinates, and men's surveys. Rutstein and Johnson's 2004 Comparative Report formalised the asset-based wealth index, and the periodically updated Guide to DHS Statistics codified indicator definitions, making the programme the principal source of comparable population and health data for the developing world.
When to use it
Use DHS data and methods when you need nationally representative, internationally comparable indicators of fertility, mortality, family planning, maternal and child health, and nutrition, especially in settings lacking complete vital registration. They are ideal for monitoring health trends, benchmarking countries, and analysing socio-economic and geographic inequalities in health. They are less suited to measuring rare events in small subpopulations, to fine-grained small-area estimates, or to detecting short-term programme effects, since surveys run only every few years and cross-sectional designs limit causal inference.
Limitations and considerations
- Cross-sectional design limits causal inference and cannot directly attribute changes to specific programmes.
- Retrospective birth and mortality histories are subject to recall error, age heaping, and the omission or displacement of births and deaths.
- Sample sizes support national and regional estimates but not reliable district or small-area figures without modelling.
- Surveys are conducted only every few years, so they cannot track rapid or short-term changes in health indicators.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Standardised model questionnaires and methods yield indicators that are directly comparable across countries and over several decades.
- Birth and pregnancy histories allow estimation of fertility and child mortality without a functioning vital registration system.
- The asset-based wealth index provides a robust welfare proxy where reliable income or consumption data are unavailable.
- Publicly available, well-documented microdata and biomarkers support a vast secondary-research and equity-analysis literature.
Applications
- Tracking national and regional trends in total fertility, contraceptive prevalence, and child and maternal mortality.
- Monitoring child malnutrition (stunting, wasting, underweight) against the WHO Child Growth Standards.
- Analysing socio-economic and urban-rural inequalities in health using the wealth index quintiles.
- Estimating coverage of interventions such as immunisation, antenatal care, skilled birth attendance, and insecticide-treated nets.
- Providing inputs for global monitoring of Sustainable Development Goal health and population targets.
How it works
How it works
Data are collected with standardised model questionnaires — a household questionnaire that lists members and records assets and housing, and individual questionnaires for women of reproductive age (and often men) covering reproductive history, contraception, antenatal and delivery care, child immunisation and illness, and anthropometry. Core questions are held constant across countries and survey rounds (phases), with optional modules added to the standard core, so that indicators are comparable across space and time.
A two-stage stratified cluster sample is used: enumeration areas (clusters) are selected with probability proportional to size within explicit geographic and urban-rural strata, then households are listed and a fixed number selected per cluster. The design is deliberately complex — stratified, clustered, and often with unequal selection across domains — which must be accounted for in every estimate rather than treated as a simple random sample.
Sampling weights are computed as the inverse of the household and individual selection probabilities, adjusted for non-response and normalised, and separate weights exist for the household, women's, men's, and biomarker files. Estimation of means, proportions, and rates applies these weights and accounts for stratification and clustering so that point estimates are unbiased and standard errors and design effects correctly reflect the complex design.
Standardised indicators are then derived following the Guide to DHS Statistics: total fertility rates from birth histories, under-five and infant mortality from the full birth history using synthetic-cohort methods, contraceptive prevalence, anthropometric z-scores (stunting, wasting, underweight) against the WHO Child Growth Standards, and the asset-based wealth index built by principal component analysis and split into quintiles for equity analysis.
Common pitfalls
- Analysing DHS data as a simple random sample, which ignores weights, stratification, and clustering and understates standard errors.
- Using the wrong weight for an analysis (e.g., applying the household weight to a women's-file estimate) and biasing the results.
- Interpreting the wealth index as an absolute measure of poverty rather than a within-survey relative ranking that is not comparable across surveys without rebasing.
- Comparing anthropometric figures computed against the old NCHS reference with newer estimates based on the WHO Child Growth Standards.
Frequently asked questions
How does the DHS build its wealth index?
The wealth index, formalised by Rutstein and Johnson (2004), is constructed from household ownership of durable assets and dwelling characteristics (e.g., radio, television, bicycle, type of water source, sanitation, floor material). Principal component analysis assigns weights to these variables, the first principal component gives each household a score, and households are ranked and divided into wealth quintiles. It is a relative measure of standing within that survey, not an absolute poverty line, and is used mainly to study equity gradients in health.
Why must DHS analysis account for the complex sample design?
DHS samples are stratified, clustered, and weighted, not simple random samples. Ignoring this produces biased point estimates (if weights are omitted) and understated standard errors and confidence intervals (if clustering and stratification are ignored), because households in the same cluster are correlated. Correct analysis applies the appropriate sampling weight and declares the survey design so that variance estimation reflects strata, clusters (PSUs), and the design effect.
How are child mortality rates estimated from a DHS?
Each interviewed woman gives a complete birth history listing every live birth, its date, survival status, and age at death if deceased. From these histories, the programme computes neonatal, infant, and under-five mortality for recent calendar periods using a synthetic-cohort life-table approach that combines age-specific mortality probabilities. This reconstructs mortality levels and trends for the years preceding the survey without relying on civil registration.
Related methods
Sources
Croft, T. N., Marshall, A. M. J., Allen, C. K., et al. (2018). Guide to DHS Statistics: DHS-7. Rockville, MD: ICF, The DHS Program.
Open publication referenceCopied.Rutstein, S. O., & Johnson, K. (2004). The DHS Wealth Index. DHS Comparative Reports No. 6. Calverton, MD: ORC Macro.
Open publication referenceCopied.