Kisoo Kim

Partisan voters tend to seek political news from media sources that match their predispositions. Scholars and pundits often attribute this partisan media sorting to psychological biases, and they typically assume that it leads voters to make worse decisions at the ballot box. To reinterpret this evidence and provide an alternative explanation, we develop two formal models of media choice—one in which voters only want to hear good news about their party and another in which voters only care about making good electoral decisions. Both models predict partisan media sorting, so sorting does not constitute evidence that voters are poorly informed or that they are driven by psychological biases. However, the models do produce competing predictions about when voters will consume more or less news and about whether signals from the news should influence vote choices. Reassessing the empirical literature, we find some support for both explanations.

How do electoral incentives motivate representatives to exert effort and shape policy positions? While reelection is widely recognized as representatives' primary motivation in democratic politics, accurately estimating its effects is challenging because representatives self-select into rerunning or retiring decisions. Leveraging legislators' lame duck status after primary elections, this article estimates how electoral incentives affect U.S. House Representatives. Compiling data from 1954 to 2022, I compare legislators who win primaries and remain motivated for reelection to those who lose and exit the race. Through regression discontinuity and difference-in-differences designs, I find that primary losers significantly reduce effort-related activities, as evidenced by decreased roll call participation, bill sponsorship, and voter communication. In contrast, the results show that legislators do not adjust policy positions for reelection; both winners and losers maintain consistent ideological and party-line adherence. These findings illuminate the scope of electoral incentives' effects and contribute to debates on electoral accountability and democratic representation.

How does a new medium affect elections and representation when it reaches citizens the prior dominant medium excluded? I study radio from 1920 to 1942, when its audio format bypassed print's literacy barrier, and instrument county radio ownership with ground conductivity and woodland share, which govern AM reception. Outside the South, a 10-point rise in radio ownership raised the Democratic two-party presidential vote share by 2.5 points within states and lowered turnout by 3.2 points; in the average county, Republican votes fell by about a tenth, most steeply where illiteracy was highest, while Democratic votes did not change. This pattern suggests that radio showed habitual Republican voters that their party no longer served their policy interests, and many stopped voting. The partisan effects vanish in the one-party South; in high-illiteracy districts outside it, Democratic seats passed to more liberal Democrats. Through its format, a new medium can change who votes, and with it election outcomes and representation.

Does the Electoral College foster candidate polarization? I develop a spatial competition model with policy-motivated candidates and endogenous turnout to show that the Electoral College polarizes candidate positions by muting the centripetal force of opposition demobilization. Under a national popular vote, platform moderation reduces turnout among opposition voters in safe states, increasing the electoral return to centrist positioning. Under the Electoral College, safe-state margins are electorally irrelevant, weakening this centripetal force. While standard models without turnout yield no clear ranking of electoral systems, I show that with endogenous turnout the popular vote always produces strictly less equilibrium divergence. The moderation gap increases with the number of uncompetitive states and, under independent shocks, scales with √n.

Empirical studies of media bias typically assume that partisan outlets always slant coverage to favor their ideological allies, so that a null correlation between owner ideology and observed bias rules out supply-side influence. I challenge this assumption through a formal model in which a politically motivated outlet chooses its editorial slant to maximize its preferred candidate's electoral prospects. The model shows that a partisan outlet praises its ally and attacks the opponent only when the race is sufficiently close: when its candidate trails badly, the outlet turns critical of its own side, and when its candidate is comfortably ahead, it turns critical or goes silent. Consequently, the same owner rationally adopts different slants across markets with different competitive landscapes, and aggregate owner–slant correlations become uninformative about the presence of supply-side effects. I contrast the distinct testable predictions of demand-driven, sincere supply-side, and strategic supply-side bias and examine them against existing data on newspaper slant and media acquisitions. These patterns illustrate that standard monotonicity-based designs cannot separate the competing mechanisms; identifying the sources of media bias requires observing how slant responds to ownership changes across the full range of local electoral conditions.

Models of electoral accountability predict that legislators should respond strategically to electoral threats. I ask whether U.S. House incumbents change their legislative behavior when an ideologically extreme challenger wins the opposing party's nomination, becoming their general-election opponent. Using a regression discontinuity design in close opposing-party primaries and constructing pre-primary ideological scores, I compare incumbents who narrowly face extreme rather than more moderate general-election opponents. I find no substantial post-primary response in party unity, ideological extremity, votes with party leaders, roll-call participation, or bill sponsorship. The estimates are consistently small across specifications and outcome measures, in line with stable legislative commitments rather than strategic adjustment in either ideological direction. Combined with existing evidence that incumbents respond to same-party primary threats, the results raise the possibility that electoral accountability operates selectively: legislative behavior responds to some forms of electoral pressure but not to variation in opposing-party challenger ideology.

Majority-party legislators are known to be more effective lawmakers than their minority-party colleagues; where that advantage is made is not. Cox-McCubbins cartel theory locates it in the majority's intrachamber agenda-control powers, and Anzia and Jackman derive its cross-sectional implication: the advantage should scale with how many of those powers a chamber's rules codify. An alternative locates it across the branches, in the co-partisan second chamber and governor that unified government supplies. Two regression discontinuity designs on a 1987–2024 panel of U.S. state legislators test both at the member level. Chamber-majority status raises a legislator's effectiveness score by roughly three-quarters of a standard deviation of that score across legislators, and the premium scales with the cartel-power index net of professionalism. A parallel design on unified state government finds that co-partisan control of the other chamber and the governorship adds no detectable premium beyond the member's own chamber majority. The majority advantage is made inside the chamber.

Broadcast television told postwar American voters almost nothing about their own representative—yet it reshaped House elections. I document that television created accountability without selection. Exploiting the staggered rollout of television across the United States in 1948–1964, I find that each year of cumulative exposure raised majority-party House incumbents' probability of returning to the next Congress by 1.6 percentage points; minority incumbents gained nothing. The premium did not reward individual records. It was unrelated to legislative effectiveness once majority status is controlled, flat across seniority, and attached to majority status itself—following legislators into and out of the majority. Consistent with collective rather than individual attribution, it was largest precisely when Congress produced least. The mechanism is informational substitution: television replaced candidate-level with party-level knowledge, shifting the House vote onto the majority party's brand. Local journalism's contemporary displacement by national digital media implies the same substitution.

Whether organized political giving buys legislative access or expresses policy preferences has framed three decades of debate. I apply a sharp regression discontinuity to the win-loss margin from the donor's perspective, using 1.7 million committee-by-race observations of U.S. House elections, 1990–2022, and extend the design to individual donors. When a committee's previously supported candidate narrowly loses, the committee redirects giving toward the winner, curtails support for the loser, leaves total giving unchanged, and reorients its partisan composition toward whichever party wins—the joint signature of access-seeking. On the subsample of committees sponsored by publicly traded firms, where lobbying is observable, the redirection is matched by a parallel, winner-specific jump in directed lobbying. Ideological committees follow the winner at least as strongly as business committees; small individual donors follow weakly and do not realign their partisan giving. Winner-following is a property of strategic donors, not of corporate status.

Does local television make representatives more accountable? The question presumes a single answer, but accountability runs through distinct evaluative channels, and congruent coverage converts information into accountability only under two conditions: local television must carry representative-specific content on that dimension, and the voter's partisanship must leave that content something to add. Using Cooperative Election Study respondents from 2006 to 2022 and redistricting-induced variation in district–media-market congruence, I find no single average effect of congruence on representative approval, but a sharp sorting across channels. Congruence sharpens ideological evaluation, which television conveys and which the party label leaves underspecified for co-partisans; it channels evaluation of district-targeted spending through partisanship, so cross-partisans punish where co-partisans do not; and it leaves four valence channels—unemployment, crime, legislative effectiveness, scandal—dormant, where television supplies no representative-specific basis for accountability. Media-enabled accountability is selective by construction: congruence determines which dimensions of representation voters police; it does not strengthen accountability uniformly.

Do corporate campaign contributions purchase measurable government benefits for donor firms? Using a regression-discontinuity design on 1,609 publicly traded firms linked through corporate PAC contributions to 588 close U.S. House general elections (2000–2022), we test four government-benefit outcomes and three political-behavior outcomes, each against a matched pre-trend placebo. Six of seven realized-outcome channels show no detectable effect; on grants we find suggestive evidence of a small positive effect with a borderline placebo. The stock-market reaction to a donor-favored win, an order of magnitude smaller than prior estimates in the literature, fails a standard bandwidth-and-polynomial robustness grid, and even under its most generous specification does not persist beyond short horizons. Corporate bond spreads show no compression for donor-connected winners either. Firms themselves do not increase lobbying, shift toward the winner's committee jurisdictions, or concentrate PAC contributions on the newly connected legislator. Our paper joins a growing null-findings literature on corporate campaign-contribution returns.

Legislative term limits are widely seen as costly to legislative capacity, yet their aggregate consequences for lawmaking remain poorly understood. I argue that the aggregate null effect of term limits on legislative productivity masks two individually significant but offsetting forces: empowered freshmen and disengaged lame ducks. Using State Legislative Effectiveness Scores across 49 states from 1993 to 2017, I show that freshmen in term-limited chambers are more effective than comparable freshmen elsewhere, while legislators in their final allowable term are less effective—consistent with the shirking documented by Fouirnaies (2022). These effects approximately cancel, leaving total legislative output unchanged. But the distribution shifts: the seniority gap in effectiveness narrows substantially, as the returns to legislative tenure decline at every level. This decomposition—which answers a question explicitly left open by Fouirnaies (2022)—reveals that term limits' costs to institutional expertise coexist with a benefit the existing literature has overlooked: a more egalitarian distribution of lawmaking success in which newcomers participate meaningfully from the start.

Does a watchful local news make legislators focus more on effective lawmaking? We construct a district-congress panel linking local daily newspapers to legislative effectiveness scores (LES) for every U.S. House member from 1873 to 2024. Newspaper presence substantially raises LES, but the effect concentrates among non-Southern Democrats before 1920: a local daily lifts a median legislator to roughly the 68th percentile of effectiveness. For all other groups, the effect is null across 104 years through 2024. Before 1920, Republican access to party machine networks substituted for the newspaper channel, and after 1920 the rise of radio reshaped the information environment. Five mechanism tests (competitiveness, aligned versus opposition press, LES decomposition, seniority persistence, and electoral consequences) favor an information-provision channel over voter accountability: newspapers help legislators learn about their districts rather than discipline them. The value of local news for legislative representation operates on the legislator side, not just the voter side.

The evidence for negative agenda control (the majority party's power to keep bills off the floor) has always been read from the bills that reached a vote, yet the power itself operates on the bills that never did. A text engine trained on three decades of House roll calls reads each never-voted bill's would-be floor position from its words and computes the probability it would have rolled the majority. The bills the majority never let reach a vote carried several times the roll risk of the bills it did, and that risk falls gate by gate from committee to the floor: the profile of sequential partisan screening, which majoritarian accounts of floor access do not predict. Blocking concentrates on broadly cosponsored coalition threats and on the social, labor, and distributive domains that split the majority: a policy map of party power that no floor-based measure can draw.