How we measure

Methodology is not an appendix at the back of a report; it is what makes a number a number. This page explains how every value we publish is built — and what we refuse to publish.

Published methodologysample N
margin ±
fieldwork date

This stamp sits on every chart we publish. If you see a number bearing our name without it, it is not ours.

📄 The Wave 1 protocol is published with a permanent digital identifier — before fieldwork, not after: doi.org/10.5281/zenodo.22559474 · all versions. An instrument disclosed after the result cannot be told apart from one fitted to it.

A number’s journey from question to publication

A methodology sheet before fieldworkEvery indicator has a fixed definition, unit, population and sampling design, approved and published before fieldwork begins. Question wording never changes after the first wave, so the series never breaks.
Stratified, multi-stage cluster samplingThe fourteen governorates as strata, districts as clusters, households at random, and the respondent within the household by the Kish grid. Weighting by population, sex and age group.
Mixed-mode fieldworkFace-to-face interviews where digital coverage is weak, telephone in the cities. Every interview is logged with its time and approximate location, and a sample of respondents is re-contacted for verification.
A published margin of error for every figureNo percentage is released without its sample size and margin of error at 95% confidence. For subgroups under two hundred respondents we report direction only, never percentages.
Independent review before releaseAn external scientific committee audits the instrument, the sample and the weighting, and its report is published with the wave. Any challenge to a result triggers a public review.
Full publicationThe questionnaire, governorate-level aggregated data and the weighting table are published with every wave under an open license with attribution.

What we commit to publicly

Standards	We adopt the principles of the AAPOR Transparency Initiative, the ESOMAR code and the WAPOR charter, and build our procedures on ISO 20252 for market and opinion research in preparation for certification.
National sample	About 2,000 respondents per wave — a design-stage margin of ±2.2% at 95% confidence, before the design effect; the actual margin is published with every wave — raised to 3,600 when subgroup precision is needed.
Frequency	The Barometer is quarterly; the economic index is monthly.
Comparative references	We compare only with those who publish their full methodology: the Arab Opinion Index (3,690 respondents over 25 days, summer 2025), the Syrian Center for Policy Research, Gallup and Pew; where no such reference exists we cite “Syria in Transition” as a non-representative guide.
What we refuse to publish	Any number without a sample, margin and fieldwork date; any comparison between differently worded questions; any data identifying an individual.
Privacy	No names or phone numbers are kept after verification ends. Data is published in aggregate only.
Independence	The launch is self-funded; funding sources are disclosed annually, and no single funder may exceed 40% of any one year’s budget (Article 46).

The sampling frame — whom we ask, and how we choose them

A country’s opinion is not measured by whoever is easiest to reach. This page states exactly whom our number represents, whom it does not, and where we fall short.

Target population	Adults aged 18 and over residing inside Syria at the time of fieldwork
Deliberately outside the frame	Syrians outside the country (measured with a separate instrument that is never merged into the national figure), and people living in closed institutions
Population frame	not yet fixed — No recent census is publicly available; the frame is therefore built on estimates that are named with their source and year, and the effect of that on weighting is published.
Fieldwork mode	Mixed mode: face-to-face where digital coverage is weak, telephone in urban centres

Multi-stage stratified cluster sample

Stratum: governorate (14) × settlement type (urban/rural)
Stage 1: sub-districts selected as clusters with probability proportional to size (PPS)
Stage 2: random start points inside the cluster, with a fixed walk rule for selecting households
Stage 3: the individual within the household by Kish grid

Weighting: Governorate, Urban/rural, Sex, Age band, Education — Design weights, then rim weighting (raking) on the variables above; weighting tables are published with every wave

Coverage map — where we measure, and under what arrangement

Full coverage — Resident team and routine field accessCoverage by prior arrangement — Local permissions and coordination are required before each round; the time each arrangement takes is documented in the wave reportPartial coverage — Some sub-districts cannot be reached; we state what was substituted and how, and compute the effect on representativeness

Published administrative boundaries (geoBoundaries), not lines of control: colour states the level of field coverage and nothing else, and the figure is the governorate’s sample in the wave. This map claims no authority, no sphere of influence and no political representation — all it says is where we measure, under what arrangement, and with how many respondents.

Sample allocation by governorate

Allocation rule: Proportional to adult population, with a floor of 80 respondents per governorate so that a valid regional estimate remains possible

Governorate	Sample	Clusters	Urban/rural %	Coverage
Aleppo	300	25	65/35	Full coverage
Rif Dimashq	240	20	55/45	Full coverage
Damascus	210	18	100/0	Full coverage
Homs	170	14	60/40	Full coverage
Idlib	160	13	50/50	Coverage by prior arrangement
Hama	150	12	55/45	Full coverage
Al-Hasakah	130	11	50/50	Coverage by prior arrangement
Deir ez-Zor	120	10	45/55	Coverage by prior arrangement
Latakia	120	10	65/35	Full coverage
Daraa	110	9	45/55	Full coverage
Raqqa	90	8	50/50	Coverage by prior arrangement
Tartus	90	8	60/40	Full coverage
As-Suwayda	80	7	55/45	Full coverage
Quneitra	30	3	40/60	Partial coverage
Total	2,000	168	—	—

This is a design proposal, fixed against the final frame before fieldwork; the adopted version is published with the wave. We show it now so it can be criticised before it is executed, not after.

Response rate — published even when it embarrasses us

Standard	AAPOR Standard Definitions (10th ed.)
We publish	RR1, RR3, CON2, REF2 — The response rate is published using AAPOR definitions with every wave, together with the full disposition table (complete · refusal · non-contact · ineligible).
If it falls	If RR3 falls below 25% we say so at the head of the report, not in a footnote, and explain what we did about it (call-backs, changed timing, instrument review).

Cognitive pretest — we test the question on people before we test people with the question

Method	Cognitive interviewing: think-aloud protocol plus retrospective probing
Interviews	24
Spread	Eight sites across four governorates differing in dialect and settlement type, half women, a third under thirty
What we look for	The question understood as intended, not as written Words that carry a prior judgement or frighten the respondent Answer scales that do not survive translation into everyday speech Questions understood in one governorate and not in another
Output	A pretest report published with the instrument: what changed and why — the question before and after
Timing	At least four weeks before fieldwork, and repeated for any question added later

A fixed core and a rotating module — why not every indicator appears in every wave

An interview past thirty-five minutes loses quality at its end, and the last figure becomes worse than its absence. So instead of touching every indicator lightly each quarter, nine are fixed and never moved, and the rest rotate — asked in depth when their turn comes.

The fixed core — every wave (9)

DIR-01
TRU-01
PRI-01
SEC-01
CPI-01
POV-01
EMP-01
FRE-01
COR-01

These build the time series, and no question wording in them changes after the first wave: changing the wording breaks the series.

Rotating module A (waves 1 · 4 · 7) (4)

SRV-01
TRJ-01
MIG-01
MIS-01

Rotating module B (waves 2 · 5 · 8) (4)

RET-01
MED-01
ELE-01
DOC-01

Rotating module C (waves 3 · 6 · 9) (3)

REC-01
PSY-01
ERW-01

The first wave carries 13 indicators

And freedom to speak is not like the others: if it falls, the trust and direction figures are read with a caveat we state in the head of the report. It measures the honesty of our own numbers before it measures the country.

Field safety protocol — published, not internal

A safety officer reporting to the ethics committee, not to field operations — so no one can press them to complete a quota

Rules for researchers

No quota outranks safety: any researcher may stop work at a site without prior permission and without loss of pay
Work in pairs at sites classed as high-sensitivity, and in daylight hours only
Check-in and check-out logged for every cluster, with an open line throughout the round
No paper forms carrying personal data; entry is digital and encrypted, uploaded and wiped from the device
Immediate-stop protocol: the round is suspended at a site on any sign of danger and resumes only by decision of the ethics committee
We never ask about party or armed affiliation, never about named individuals, and never request an ID number
Informed consent is spoken and its acknowledgement recorded; a respondent may stop at any moment without giving a reason

Respondent protection

No names or phone numbers are kept after verification ends; the deletion deadline is 90 days from the close of the wave
Coordinates are rounded to cluster level; no point that identifies a dwelling is ever published
No subgroup with fewer than 100 respondents is published, so that no individual can be inferred

Incidents: Every incident is logged, and the number and type of incidents is published in the wave report — never their locations or the names of those involved

Training: A three-day course before every wave, a third of it on safety and ethics; no one who fails it enters the field

Three lessons we learned from others before starting

We track phenomena, never people

In 2019 a large regional center published a report naming journalists, drawing criminal complaints and lasting reputational damage. We monitor phenomena — never people.

Transparency is necessary, not sufficient

A venerable institute has disclosed its donors since 1983 and has still suffered two crises. So we added a diverse board of trustees and an independent review committee.

One donor is enough to sink you

The early-2025 US aid freeze halted, according to published reports, most programs of organizations that depended on a single donor — within weeks. No source exceeds half our budget.