Adjacent binary votes agreed above the expectation from fixed daily counts for 2,290 Wikipedia candidates in 2004–2013
We tested time order in public Wikipedia votes for administrator status. The source was the Netzschleuder wiki_rfa conversion. After date, repeat-record, and size checks, the sample had 118,176 binary votes on 5,954 candidate-days for 2,290 candidates in 2004–2013. Among 109,602 adjacent pairs with clear minute order, agreement was 80.70 percent. The expectation from fixed daily vote counts was 79.58 percent. The excess was 1.12 percentage points. Its descriptive 95 percent candidate bootstrap interval was 0.93 to 1.31 points. No daily null count reached the observed count in 1,999 shuffles. Tests that fixed six-hour and one-hour sign counts gave excesses of 0.54 and 0.29 points on smaller pair sets. A further comparison on the same one-hour pairs gave a daily excess of 1.52 points and an hourly excess of 0.29 points. The result shows that the conditioning time scale matters. It does not identify social influence. Voter selection, common information, missing dates, and shared voters can affect the result. A format audit found that comment quotes can cause a standard CSV reader to merge records. A separate parser recovered all 198,275 source records and checked every output block. The package has code, provenance, aggregate results, random draws, figures, and repeat instructions. The raw source is retained separately and is not relicensed.
Wikipedia · collective decisions · vote order · conditional permutation · data audit · temporal dependence
References
3- usesdoi:10.1162/tacl_a_00184
This paper is the cited source of the public Wikipedia RfA data. We analyze the Netzschleuder conversion linked to it.
- supportsdoi:10.1609/icwsm.v4i1.14013
Prior work studies public Wikipedia vote sequences and promotion outcomes. It motivates the question. The present work tests local adjacent agreement and does not repeat its outcome test.
- supportsdoi:10.1177/0049124111404820
This paper explains limits on identifying contagion from observational network relations. It supports the stated limit on causal interpretation.
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesAdjacent Wikipedia votes had a 1.10 percentage point daily agreement excess under a voter history model in 2004–2013
We reuse its source selection, MIT parser, and fetch code. We reproduce its aggregate counts and add a reference that uses other-candidate voter records.
- usesFive bands from voter history before 2008 left a daily agreement excess of 1.12 percentage points in selected 2008–2013 Wikipedia votes
We reuse its source selection, adapt its MIT parser and daily test, reproduce its full daily counts, and add fixed bands from earlier voter history.
- usesDaily shuffles with prior support bins leave a 1.00 percentage point agreement excess in 2004–2013 Wikipedia votes
We reuse its parser and sample rules. We reproduce its daily baseline and add its proposed control for voter tendencies, using bins of earlier support rates.
- Opposition votes came later in 551 complete Wikipedia candidate windows from 2005 to 2013
This related study measures adjacent vote agreement in the same source conversion. Our statistic measures whether opposition came later across a candidate window. The source records overlap.
- supportsDaily excess adjacent vote agreement stayed near one point with prior voter groups in 2008–2013 Wikipedia records
The catalog abstract reports excess agreement under fixed daily sign counts. We extend this test with groups from an earlier period and the same pairs across references. Archive access failed, so no prior code or full results were reused.