Standard Eurobarometer, spring rounds 2012–2026 (country aggregates)标准欧盟晴雨表 2012–2026 年历年春季轮（国家层面聚合）

The Eurobarometer has surveyed public opinion across Europe since 1973. This dataset is a selected subset of that programme, not the whole series: it takes one round per year — the spring round of the Standard Eurobarometer, the one the European Commission publishes as "Standard Eurobarometer NN – Spring YYYY" — for 2012–2026. One row is one country in one year.

Rounds included: EB 77.3 (ZA5612, 2012): 33 countries; EB 79.3 (ZA5689, 2013): 33 countries; EB 81.4 (ZA5928, 2014): 33 countries; EB 83.3 (ZA5998, 2015): 33 countries; EB 85.2 (ZA6694, 2016): 33 countries; EB 87.3 (ZA6863, 2017): 33 countries; EB 89.1 (ZA6963, 2018): 33 countries; EB 91.5 (ZA7576, 2019): 33 countries; EB 93.1 (ZA7649, 2020): 33 countries; EB 95.3 (ZA7783, 2021): 38 countries; EB 97.5 (ZA7902, 2022): 38 countries; EB 99.4 (ZA7997, 2023): 38 countries; EB 101.3 (ZA8843, 2024): 37 countries; EB 103.3 (ZA9128, 2025): 37 countries; EB 105.2 (ZA9144, 2026): 37 countries.

Three items are hosted: satisfaction with the way democracy works in the respondent's own country, holding a positive image of the European Union, and tending to trust the European Union. Figures are weighted shares computed with each round's national weighting factor.

Two things to keep in mind when reading the numbers. First, don't-know answers are excluded from the base rather than imputed, while the European Commission's own published percentages keep them in it, so the figures here run a few points higher than the Commission reports — German trust in the EU in 2017, for example, is 50.6% here and 46.5% on the Commission's base. Second, an item a round did not field in a given country is left empty rather than filled: satisfied_democracy: not fielded in ALB, HRV, ISL, MKD, MNE, SRB, TUR in 2012-2019 (41 empty cells).

The variable names for these items change from round to round (qa17a/qa9/qa8a_8 in the 2010s, sd18a/d78/qa6_x from 2020), so the cleaning script locates them by variable label and derives the response codes from the value labels rather than hard-coding question numbers. Two traps in the source files are handled explicitly: several rounds carry a second variable with an identical label that is valid only for the Turkish Cypriot community sample, which is skipped; and in the 2021 and 2022 rounds the fourth response category of the democracy-satisfaction item is mislabelled as a don't-know in the file, which GESIS records as an erratum, so those two rounds use the corrected coding.

East and West German samples are combined with the official combined weight, as are Great Britain and Northern Ireland where both are fielded; the Turkish Cypriot community sample (CY-TCC) has no ISO3 code and is a separate sample from Cyprus, so it is dropped.

欧盟晴雨表（Eurobarometer）自 1973 年起持续调查欧洲公众舆论。本数据集是该项目的**选定轮次子集**，不是全序列：每年只取一轮——标准欧盟晴雨表的春季轮，即欧委会对外发布的"Standard Eurobarometer NN – Spring YYYY"——覆盖 2012–2026 年。一行 = 一个国家在某一年。

纳入的轮次：EB 77.3（ZA5612，2012）：33 个国家/地区；EB 79.3（ZA5689，2013）：33 个国家/地区；EB 81.4（ZA5928，2014）：33 个国家/地区；EB 83.3（ZA5998，2015）：33 个国家/地区；EB 85.2（ZA6694，2016）：33 个国家/地区；EB 87.3（ZA6863，2017）：33 个国家/地区；EB 89.1（ZA6963，2018）：33 个国家/地区；EB 91.5（ZA7576，2019）：33 个国家/地区；EB 93.1（ZA7649，2020）：33 个国家/地区；EB 95.3（ZA7783，2021）：38 个国家/地区；EB 97.5（ZA7902，2022）：38 个国家/地区；EB 99.4（ZA7997，2023）：38 个国家/地区；EB 101.3（ZA8843，2024）：37 个国家/地区；EB 103.3（ZA9128，2025）：37 个国家/地区；EB 105.2（ZA9144，2026）：37 个国家/地区。

托管 3 个题目：对本国民主运行的满意度、对欧盟印象正面、倾向信任欧盟。数值为使用各轮国内权重计算的加权比例。

读数时有两点要注意。其一，不知道/拒答从分母剔除、不做插补，而欧委会公布的百分比是把它们计入分母的，因此本表数值会比欧委会报告高几个百分点——例如德国 2017 年的信任欧盟，本表 50.6%，按欧委会口径是 46.5%。其二，某轮在某国没问的题目留空、不填补：satisfied_democracy：2012-2019 年在 ALB、HRV、ISL、MKD、MNE、SRB、TUR 未问（41 个空格）。

这三道题的变量名逐轮会变（2010 年代是 qa17a/qa9/qa8a_8，2020 年后是 sd18a/d78/qa6_x），因此清洗脚本按**变量标签**定位题目、并从**值标签**推出命中码，不写死题号。原始文件里有两个坑已显式处理：多轮存在标签完全相同、但只对北塞浦路斯土族社区样本有效的平行变量，已跳过；2021 与 2022 两轮的民主满意度题第 4 档在文件里被误标成「不知道」（GESIS 已列为勘误），这两轮按勘误后的编码计算。

德国东西部样本按官方合并权重合并；英国本土与北爱尔兰若分开抽样也同样合并；北塞浦路斯土族社区样本（CY-TCC）没有对应 ISO3、且与塞浦路斯是两个独立样本，已剔除。