Laboratory for Electoral
Systems and Technologies

Electoral Clientelism (2014–2024)

How widespread vote buying and coercion are, how they vary across election types and parties, and why machine voting and video surveillance fail to curb them.

• The share of controlled and bought votes ranges from 8.8% (presidential, November 2021) to 29.1% (parliamentary, October 2014); the trend is U-shaped — a decline around the 2020–2021 protests and a rebound above 20% (parliamentary, April 2023: 27.9%; local, October 2023: 28.9%) despite the introduction of voting machines and video surveillance.
• Cash offers dominate over threats (2014: 81.5% cash versus 9.4% threats), and the average price per vote has risen with inflation — from about BGN 39 (2014) to BGN 80–87 (2022–2023).
• Field-wave estimates put directly bought and controlled votes at 180,000–240,000; including households and paid turnout across roughly 12,000 precincts, the figure rises to over 1 million (nearly one-third of votes cast), worth on the order of BGN 100 million.

Six nationally representative field surveys (N from 936 to 1,699), 2014–2024 · model "Share = Offered × Accepted × Delivered × Turnout" · working manuscript

Campaign Finance in the EU (2012–2024)

How the form of government (semi-presidential or parliamentary), the electoral system, and the financing model shape the relationship between campaign spending and electoral outcomes.

• Each additional percentage point of spending yields a different return across countries: +1.04% in France (R²=0.43), +5.0% in Romania (R²=0.64), and, most strikingly, +27.9% in Austria (R²=0.54); the tightest fit is in Croatia (R²=0.86).
• Public financing strengthens the spending–outcome relationship the most (coefficient 5.76), while predominantly private financing is statistically insignificant; the three institutional factors together explain only 17.5% of the variance in outcomes.
• A comparison of all 27 EU member states (cluster analysis, four groups; ANOVA F=11.06, p<0.001) shows that the most tightly regulated systems have lower turnout and weaker party institutionalization — a "moderate" level of regulation appears optimal.

Original spending and outcome data (13 countries) and a comparative analysis of all 27 EU member states · correlation, regression, and cluster analysis · working manuscripts

Election Administration in Bulgaria

How election administration — legal framework, technology, precinct committees, oversight — affects turnout and trust, seen through the eyes of the administrators themselves.

• Twelve in-depth interviews with election administrators (2024) reveal growing complexity: the Electoral Code has been amended more than 20 times since 2014, and preferential voting is a leading source of counting errors.
• Machine voting produces "clean" protocols when it is the sole method (July 2021), but under a mixed system the machine functions de facto merely as a printer and slows down the count; reliance on a single vendor constitutes a systemic risk.
• Precinct election commissions are the weakest link (an aging, low-paid membership, with no sanctions for systemic errors); trust in the fairness of elections stands at 10% (2024), and turnout has fallen from 49.1% (April 2021) to 33.4% (June 2024).

12 semi-structured interviews (2024) and official CEC turnout aggregates · thematic analysis · working manuscript

European Parliament Elections: The Second-Order Elections Model (1999–2024)

Whether the "second-order" character of EP elections holds up empirically across the entire EU, or whether the 2024 data point to an erosion of this logic.

• Across 153 observations (country-years; the 27 EU member states plus the United Kingdom), average turnout in EP elections is 47.3% versus 68.9% in the preceding national elections — a gap of −21.6 percentage points.
• Compulsory voting narrows the gap to nearly zero (−7.1 p.p. versus −24.0 without it); the three-group typology is statistically distinct (ANOVA F(2,150)=338.1, p<0.001).
• A time-series forecast (ARIMA) through 2034: "second-orderness" is slowly weakening at the aggregate level but remains deep in a subset of countries — the model holds up, but with clear limits.

Comparative panel, 153 observations, 1999–2024 · regression, typology, ANOVA, and ARIMA forecasting · late-stage working draft

Machine Voting and Voter Turnout

How the introduction of machine voting affects turnout and attitudes toward the electoral process.

• In a nationally representative survey (N=1,699, April 2024), 16.9% report that they or someone close to them had been offered money or threatened over their vote in the past three years.
• Trust in machine voting is split — 16.7% express full trust versus 22.9% full distrust; the leading concerns are the risk of hacking or manipulation (41.1%) and technical malfunctions (40.8%).
• No statistically significant relationship is found between party affiliation, education, or age and attitudes toward the machines — skepticism is distributed evenly across groups.

Nationally representative survey (N=1,699, April 2024), ±2.38% · descriptive and inferential statistical analysis · submitted for peer review

Machine Voting (Bulgarian National Science Fund, 2025)

Machine voting in the context of electoral security and democratic representation: international standards and Bulgarian practice.