Two instruments, one trend?
When two major surveys measure the same country, do they see religion changing in the same direction?
Each dot is a country’s religiosity trend as the WVS/EVS measures it (across) and as the ESS measures it (up), with 95% intervals on both; the diagonal is perfect agreement.
Attendance: 24 countries; cross-instrument r = .61, below the pre-committed .70 bar — the instruments diverge. Both confirm a fall in 10 and a rise in 2; they contradict each other in 2; 10 are unconfirmed.
- Fall confirmed by both
- Rise confirmed by both
- Instruments contradict
- Unconfirmed
- Diagonal: both instruments agree exactly
Each dot is one country, placed by the same quantity estimated separately in each instrument, with its 95% interval on both axes. On the diagonal the two agree exactly; the shaded quadrants are where they point in opposite directions. The three trend families share one frame so they can be compared; the cross-section is a different quantity and has its own. Agreement classes are published for attendance only.
Attendance: WVS/EVS and ESS estimates · 24 rows
| Country | WVS/EVS | WVS/EVS lower 95% | WVS/EVS upper 95% | WVS/EVS N | ESS | ESS lower 95% | ESS upper 95% | ESS N | Class |
|---|---|---|---|---|---|---|---|---|---|
| Bulgaria | 0.05 | -0.02 | 0.13 | 4,006 | 0.07 | 0.04 | 0.11 | 15,264 | Unconfirmed |
| Croatia | -0.12 | -0.21 | -0.03 | 2,984 | -0.28 | -0.32 | -0.23 | 8,009 | Fall confirmed by both |
| Cyprus | 0.16 | 0.1 | 0.23 | 3,995 | -0.04 | -0.09 | 0.02 | 6,724 | Unconfirmed |
| Czechia | -0.01 | -0.06 | 0.04 | 4,742 | -0.06 | -0.09 | -0.03 | 19,867 | Unconfirmed |
| Estonia | -0.12 | -0.19 | -0.05 | 4,328 | -0.11 | -0.13 | -0.08 | 18,102 | Fall confirmed by both |
| Finland | -0.13 | -0.21 | -0.06 | 3,330 | -0.1 | -0.12 | -0.07 | 21,073 | Fall confirmed by both |
| France | -0.04 | -0.1 | 0.02 | 4,363 | -0.05 | -0.08 | -0.03 | 20,745 | Unconfirmed |
| Germany | -0.03 | -0.08 | 0.02 | 9,771 | -0.16 | -0.18 | -0.14 | 36,611 | Unconfirmed |
| Greece | -0.02 | -0.11 | 0.07 | 2,685 | -0.08 | -0.1 | -0.05 | 15,261 | Unconfirmed |
| Hungary | 0.01 | -0.06 | 0.08 | 4,012 | -0.01 | -0.03 | 0.02 | 18,560 | Unconfirmed |
| Italy | -0.21 | -0.26 | -0.15 | 4,727 | -0.05 | -0.09 | -0.01 | 12,847 | Fall confirmed by both |
| Latvia | -0.14 | -0.2 | -0.08 | 2,797 | -0.17 | -0.22 | -0.12 | 5,113 | Fall confirmed by both |
| Lithuania | 0.04 | -0.03 | 0.12 | 2,916 | -0.09 | -0.15 | -0.03 | 12,885 | Unconfirmed |
| Netherlands | -0.19 | -0.23 | -0.15 | 8,723 | -0.17 | -0.19 | -0.15 | 20,002 | Fall confirmed by both |
| Norway | 0.1 | 0.03 | 0.17 | 3,228 | -0.09 | -0.12 | -0.06 | 17,377 | Instruments contradict |
| Poland | -0.2 | -0.27 | -0.13 | 4,744 | -0.23 | -0.25 | -0.21 | 18,902 | Fall confirmed by both |
| Russia | 0.13 | 0.08 | 0.19 | 9,324 | 0.25 | 0.17 | 0.32 | 12,024 | Rise confirmed by both |
| Slovakia | -0.1 | -0.16 | -0.04 | 4,089 | 0.02 | -0.01 | 0.06 | 12,626 | Unconfirmed |
| Slovenia | -0.1 | -0.17 | -0.03 | 4,483 | -0.12 | -0.15 | -0.1 | 14,642 | Fall confirmed by both |
| Spain | -0.02 | -0.09 | 0.05 | 5,039 | -0.15 | -0.18 | -0.13 | 21,175 | Unconfirmed |
Estimand: per-country weighted standardized trend slope of each religiosity item family (z within country x instrument; slope in SD/decade, 95% CI), estimated separately in WVS/EVS and ESS; plus the per-country cross-sectional attendance->left-right standardized slope from each instrument. Verdict rule: per family: instruments AGREE iff cross-country Pearson r >= 0.70 (pre-committed); < 8 overlap countries -> no test. Country classifier on rounded 95% CIs: agree_fall / agree_rise / contradict / unconfirmed. Floors: complete-case N>=400, >=2 readings spanning >=8 years per country-instrument-family regression; Kish effective N>=100 per plotted year cell; >=8 overlap countries per family verdict. Join: WVS<->ESS compared at COUNTRY level (country+wave key; no shared respondents); approximate tier; never pooled. Weighting: WVS `weight`; ESS `anweight` (pspwght*pweight for pre-R9 rounds). Cross-country agreement statistics equal-weight each country.
World Values Survey / European Values Study 2002–2023 and European Social Survey rounds 1–11 · country-level standardized slopes
Reading notes
Each slope is estimated within one instrument; the two are compared country by country and never pooled. Verdicts apply a pre-committed r ≥ .70 rule; prayer has too few overlapping countries to test. Agreement classes are published for attendance only.
Source & method
Kept the per-country trend slopes (SD per decade) for the three religiosity families — attendance, belonging, prayer — from both WVS/EVS and ESS with their 95% CIs and complete-case n, the source's agreement class for each attendance country, and the cross-sectional attendance → left–right slopes from both instruments. Kept each family's pre-committed verdict inputs (cross-instrument r, overlap count, the r ≥ 0.70 bar and the 8-country floor). Dropped the fitted trend lines, per-year series panels and dumbbell ranks. Estimates and bounds rounded from four to three decimals.
Capture provenance
- box:Politics / book2/chapters/do-the-instruments-agree/data.json
- Modified 2026-07-02 · SHA-256 18145bba2780e8a46c7ef4b3e2e2a4c42caadf69364ecc4d683df418eed16950
- Captured
- 2026-09-12
- Fixture SHA-256
- 3f5c09533856d2f08a32c5eb612f5ecc84b25666a632e0b5606404267e9ebc81
- Reuse note
- Aggregates of the World Values Survey / European Values Study (2002–2023) and the European Social Survey (ESS ERIC, rounds 1–11), computed at country level by the book's analysis pipeline; the two instruments are compared, never pooled. Both are free for non-commercial use with citation.
