This study addresses an interesting question on political participation in direct versus representative democracies. That is, are citizens more likely to turn out to vote in direct issue-oriented referendums or in indirect candidate-oriented public office elections? In order to answer this question, this study takes advantage of a rare case of a natural experiment in Taiwan’s 2018 concurrent referendum and local elections. Capitalizing on a gap between the voting-age eligibility rule of 18 for referendums and 20 for local offices, this study develops rigorous regression-discontinuity (RD) designs in causal inference with the “two cutoffs model” for the referendums and “the standard single cutoff model” for the local elections. The two RD designs are then applied to a unique set of individual-level validated turnout data with large sample size released by the Central Election Commission. Empirical estimates of the two RD designs confirm our hypothesis that, other things being equal, the first-time voters are much more likely to turn out when they are eligible for both referendum and public office ballots than when they are only eligible for referendum ballots. This finding has important implications not only for theories of direct versus indirect democracies but also in practice for democratic decision-making mechanisms.