An interactive 3-hour Practice Pack on evaluating Indian livelihoods programmes — grounded in the Sustainable Livelihoods Framework, PLFS/NRLM/SECC-aligned, with realistic 2026 ₹ budgets. Fill the forms as you go; the capstone builds itself.
4 modules~3 hoursInteractiveIndia · India-context
Your progress
0% complete
Your Capstone
One-page Livelihoods Evaluation Design Brief
Walk in with a livelihoods programme (rural, urban informal, skills — any). Walk out with a defensible design grounded in the Sustainable Livelihoods Framework — outcome dimension chosen, PLFS/NRLM-aligned instruments, household sampling plan, seasonal data calendar, mixed-methods integration, honest framing.
Module 1 · ~30 min
Which livelihood dimension are you actually evaluating?
"Livelihoods" is one of the most over-used and under-defined terms in Indian development practice. A programme officer says "we improved livelihoods" and means anything from "households earned more this year" to "women joined SHGs" to "the village got a new road." Before any evaluation design, the field needs to make explicit which dimension of livelihood you are claiming to affect — and therefore measure.
The Sustainable Livelihoods Framework (DFID 1999) — five capitals
The most useful organising lens. Livelihood outcomes flow from a household's combination of five capitals:
Human capital — skills, health, education, ability to labour
Social capital — networks, trust, group membership (SHGs, kinship)
Natural capital — land, water, forests, climate stability
Physical capital — roads, electricity, irrigation, housing, livestock
Financial capital — savings, credit, insurance, remittances
The vulnerability context (shocks, seasonality, trends) and policies/institutions/processes mediate how capitals convert into outcomes — income, food security, wellbeing, agency.
The four common outcome questions
Income / consumption — "Did household income or expenditure rise?" Most fundable claim, narrowest measure. PLFS-comparable.
Asset / capital accumulation — "Did the household's productive assets grow?" Captures durability more than income; SECC-aligned proxies.
Capability / agency — "Can the household now do what it values?" Sen-Nussbaum framing; harder to measure, more meaningful.
Vulnerability reduction — "Is the household less exposed to shocks?" Often the most defensible programme claim for poverty programmes.
Standard design dimensions still apply
The same three dimensions from any evaluation design hold for livelihoods, but with India-specific twists:
Question type: outcome vs process vs theory-based. For SHG/NRLM-type programmes, process tracing matters as much as outcome.
Evidence type: quant vs qual vs mixed. PLFS-comparable quant is essential if you want policy uptake; qual is essential if your claim is about agency or process.
Time dimension: livelihood effects compound or attenuate over years — a one-year evaluation usually under-detects effects. Plan for at least pre-post, ideally longitudinal.
Outcome dimension: Vulnerability reduction (primary) + asset accumulation (secondary). Income alone would mask the smoothing effects of SHG savings.
SLF capitals targeted: Social (SHG networks), financial (savings, credit), human (skills training). Natural and physical secondary.
Design: Theory-based mixed-methods with matched comparison. Pre-post on a sample sub-cohort + retrospective recall on the broader frame. Honest about limits.
Your SLF + Outcome Decision Sheet
Pin down which dimension(s) your programme actually claims to affect — and therefore needs to measure.
Which of the five capitals does your programme aim to build? Be honest about primary vs secondary.
"We will use this to..." — with a date.
Saved
Self-check
A drought hits the year of your endline survey. Household income in your livelihoods programme is statistically flat. The control group's income dropped 12%. What does this finding actually show?
The programme failed — no income gain
The programme produced significant vulnerability reduction, not visible if measured by income alone
The drought makes the evaluation invalid
Wait another year before reporting
Correct. This is exactly why outcome dimension choice matters. The household's resilience to the shock IS the finding — the programme protected income that would otherwise have fallen 12%. Reported as income, it looks like "no effect"; reported as vulnerability reduction, it's a substantial positive finding.
Module 2 · ~30 min
Choosing instruments that align with Indian data systems
The hardest part of livelihoods evaluation is choosing instruments that are both rigorous and also comparable to the data the rest of the Indian policy ecosystem uses. Studies that ignore PLFS/NSSO/NFHS instruments lose the ability to triangulate, calibrate, or be compared with national stats.
Recall windows matter. Income / consumption recall over 30 days is much more reliable than 1 year. Asset stock can use longer windows; income flow cannot.
Seasonality dominates. Same instrument administered in February vs August produces very different income/consumption answers. Either fix the season or measure across multiple rounds.
Household vs individual unit. Many livelihood outcomes are individual (women's earnings, women's savings) but most instruments default to household head reporting. This systematically undercounts women's economic activity.
PLFS-comparability matters operationally
Funders and government partners increasingly ask "is this PLFS-comparable?" before reading the rest of the report. Building 30-50% of your quant instrument around PLFS / HCES question formats means your findings can be benchmarked against national data — and your effect size estimates have something to be calibrated against. Lose this and your numbers float in space.
Your Instrument Selection Matrix
Pick instruments that match the outcome you chose in Module 1. PLFS-anchoring strongly recommended.
e.g., "NRLM MIS data on SHG participation + 40 life-history interviews + seasonal calendars in 8 villages"
Household or individual? If household, who answers? Note: defaulting to head-of-household undercounts women's economic activity.
Which of your modules use PLFS / HCES / NFHS / SECC question formats? Aim for 30-50%.
Saved
Self-check
Your livelihoods programme primarily builds SHG-mediated financial inclusion. Funder asks for a one-instrument summary measure of impact. What's the best choice?
Decline this framing — explain why no single instrument can summarise a multidimensional programme
A programme-developed composite SHG-impact score
Correct. The funder's framing collapses the actual programme effect (savings smoothing, credit access, social capital, women's agency) into one number that under-represents most of it. The harder, honest move is to explain that 3-4 measures jointly tell the story, and offer to draft a 1-page summary that integrates them.
Module 3 · ~30 min
Sampling, seasonal calendar & the migration complication
Livelihoods sampling has three structural quirks that SEL / education / health evaluations don't share:
The household unit is unstable. Members migrate seasonally; some return, some don't. Defining "who is in the household" requires explicit decisions.
Seasonality is not a nuisance variable — it IS the variable. Lean season vs harvest season household economics look different by orders of magnitude.
Programme participation is endogenous. Households that join SHGs are systematically different from those that don't. Naive comparisons fail; matching/IV/DiD designs are usually necessary.
Life-history interview (90 min + transcription): ₹3,500-5,500 each
FGD (90-120 min, 8-10 participants): ₹8,000-12,000 each
NRLM/SHG MIS data extraction (programme-internal): typically free; budget ₹50,000-1,00,000 for cleaning and analysis
Seasonal calendar workshop (village-level): ₹4,000-7,000 per village
The seasonal calendar question
For most livelihoods programmes, you need at least two data collection rounds — one in a lean season, one in a peak season — to characterise the household economy honestly. Single-round data systematically distorts. The cleanest designs use three rounds: baseline (lean), midline (peak), endline (lean again — same season as baseline).
Migration: the silent confound
20-40% of working-age men in many rural Indian household samples migrate for 3-9 months per year. Standard household surveys either undercount them (when they're absent) or wrongly include them (when they're present but earning elsewhere). Decide your migration coding protocol explicitly — and document it. Common approach: "usually-resident" definition with separate migration module that captures absent members' earnings.
Your Sampling Plan & Seasonal Calendar
Numbers + seasons + migration coding. These flow into your capstone.
Households per arm × number of villages/clusters. e.g., "800 households across 32 villages (16 treatment + 16 matched comparison)"
e.g., "Baseline: June-July (early monsoon, lean). Midline: October-November (post-harvest, peak). Endline: June-July (next year, same lean season as baseline)."
How will you define "household member"? How will you capture absent migrants' earnings?
Matched non-participants? Wait-listed treatment? Adjacent non-programme villages? Be honest about selection issues.
e.g., enumerator turnover, monsoon access, election interference, festival windows.
Saved
Self-check
You're designing a household survey for an NRLM-area livelihoods evaluation. Budget says you can do 600 households in ONE round, or 400 households across TWO rounds (lean + peak season). Which is the stronger design?
600 households, one round — bigger sample, more power
400 households across two rounds — captures seasonality, which is the dominant signal
600 households, one round, timed for peak season
Mix: 300 in lean + 300 in peak, different households
Correct. For livelihoods, seasonality is not noise — it's the signal. A single-round 600-HH study reports an artificially-precise number that's actually a snapshot of one season. Two-round same-households gives you seasonality + change over time, which is what livelihoods programmes actually affect.
Module 4 · ~25 min
Analysis, integration, and the multi-dimensional honesty
Livelihoods reporting has its own specific pathologies:
Single-number worship. "Income rose 18%" — clean, headline-friendly, often masks who in the household captured the gain and at what cost.
Confounding shock attribution. "Income rose during a good monsoon" — was it the programme, or the rain?
Gender-blind aggregation. Household-level data hides intra-household distribution. The programme that "raises household income" may concentrate it in male hands.
SHG-MIS-as-impact. Programme MIS data reports activity (groups formed, loans disbursed). Activity is not impact.
Analysis priorities
Disaggregate by gender always. Within-household if your design captures it; across-household at minimum.
Decompose effects by SLF capital. Where did the gains come from? Financial inclusion? Skills? Asset accumulation? The decomposition is what makes the finding redesignable.
Honest about shock attribution. Always benchmark against PLFS/NFHS regional-level trends. If your treatment group rose 12% and the regional average rose 11%, your programme effect is 1% — not 12%.
Mixed-methods integration. The qualitative life-histories tell you what changed in households; the quant tells you how widespread. Both are needed; neither is sufficient.
Appendices — full results, instruments, qualitative quotes, PLFS comparisons
The Krishna question
Anirudh Krishna's "Stages of Progress" work documents that the most consistent thing livelihoods programmes do is prevent downward mobility during shocks (illness, weather, migration failure) — not produce upward mobility. Frame your findings with this honestly. "We helped households not fall" is a real, defensible, fundable claim. "We lifted households out of poverty" usually isn't.
Your Reporting Plan & Honesty Frame
How will you report? To whom? What's your honest framing?
By gender always; by what else? Caste, age, sub-region, programme-component, SLF capital?
How will you separate programme effect from regional trends / monsoon / general policy? What national/regional comparisons will you use?
How will quant + qual strands actually integrate?
For each: format, length, key questions they'll ask, language.
"This evaluation will tell us ___ and will NOT tell us ___."
Saved
Self-check
Your livelihoods evaluation reports: "Treatment villages saw 14% household income growth, control villages 11%, regional PLFS-trend 10.5%." What's the honest programme effect?
14% income growth
3% effect (treatment minus control)
Approximately 3-3.5% effect, but the control-vs-trend gap (0.5%) suggests possible matching/selection issues to investigate
No effect — within statistical noise
Correct. The naive headline (14%) ignores the counterfactual. The treatment-vs-control difference (3%) is the programme effect, but the fact that the control rose almost identically to the regional PLFS trend (11% vs 10.5%) suggests the matching was decent. If control had outperformed regional trend significantly, you'd worry about selection bias in your matching. Always benchmark all three.
Capstone
Your Livelihoods Evaluation Design Brief
Click Build my brief to compile your module answers into a one-page artefact. Copy as markdown, print as PDF, or share with your team.
One-Page Livelihoods Evaluation Design Brief
Click "Build my brief" — your module answers will be pulled into the artefact. Edit/refine afterwards if needed (click in the box and type).
Your brief will appear here when you click "Build my brief".
It will draw from your answers in Modules 1-4 (which are saved in your browser). Empty fields show as placeholders — you can either go back and fill them, or edit them here directly after building.
Share the brief with one colleague before circulating widely. Their first reaction is the most accurate signal you'll get on whether the framing lands.