Chapter 15: Campaign Effects — Do They Matter?

Three weeks before Election Day, the Garza campaign's internal tracking poll showed Garza trailing Whitfield by two points. The margin had barely moved in six weeks despite a television advertising blitz that had consumed most of the campaign's summer budget, a bus tour through twelve counties, and two debate performances that the campaign's own post-debate focus groups rated as "strong" or "very strong" for Garza.

Jake Rourke studied the tracking numbers in the campaign war room and said what he was thinking: "The ads aren't moving the dial."

Nadia Osei looked at the same numbers and thought something slightly different: "The structural environment was always going to put us behind two points at this point. The question is whether we're ahead of where we would be without everything we've done."

This is the fundamental challenge in analyzing campaign effects: how do you know what would have happened in the counterfactual world where the campaign didn't exist, or existed differently? The tracking poll shows Garza at 48 points after massive advertising spending. Is that good or bad? It depends on what the structural baseline would have been. If the structural model said Garza should be at 45, the advertising worked brilliantly. If the structural model said she should be at 52, something went badly wrong.

The science of campaign effects has grappled with this counterfactual problem for more than seventy years, and the resulting research is both richer and more contested than most practitioners realize.

15.1 The Minimal Effects Tradition

15.1.1 Lazarsfeld and the Columbia School

The systematic study of campaign effects begins, by most accounts, with Paul Lazarsfeld's The People's Choice, published in 1944. Lazarsfeld and his colleagues Bernard Berelson and Hazel Gaudet conducted a panel survey of voters in Erie County, Ohio, over the course of the 1940 presidential election between Franklin Roosevelt and Wendell Willkie. Rather than a single cross-sectional snapshot, the panel design interviewed the same respondents multiple times through the campaign, allowing the researchers to track how individual opinions changed — or, more precisely, didn't change.

The primary finding was striking in its negativity. Few voters changed their candidate preference between the campaign's start and Election Day. Among those who did change, the most common movement was from undecided to decided — "crystallization" of a pre-existing preference rather than genuine conversion. The most dramatic finding was that campaign exposure was more likely to reinforce existing preferences than to convert opponents. Voters who supported Roosevelt tended to seek out pro-Roosevelt media; voters who supported Willkie tended to find and remember pro-Willkie messages. This pattern — which Lazarsfeld called "selective exposure" — meant that campaigns were primarily talking to the converted.

The research produced what became known as the "minimal effects" hypothesis: campaigns primarily activate and reinforce pre-existing partisan and social group loyalties rather than converting voters from one side to the other. The strong predictors of vote choice — economic conditions, presidential approval, long-standing partisan identity — mattered far more than any specific campaign activity.

💡 Intuition Check: Why Did the Columbia School Find Minimal Effects? Several methodological and substantive features of the Erie County study shaped its conclusions. Erie County in 1940 was a relatively stable, low-diversity community with deeply rooted partisan loyalties. The panel study ran through an entire election but in an era of far fewer media sources and a far smaller advertising industry. Most importantly, the study measured opinion change — a high bar. Campaigns can have significant effects on enthusiasm, turnout, issue salience, and knowledge without producing visible opinion switching. Later scholars argued the minimal effects finding was partly a measurement artifact.

15.1.2 Converse, Stability, and the Party Identification Anchor

Philip Converse's 1964 essay "The Nature of Belief Systems in Mass Publics" reinforced the minimal effects tradition through a different route. Converse used ANES panel data to show that for most Americans, political attitudes were not organized into coherent ideological structures but were largely non-attitudes — responses that fluctuated quasi-randomly across survey waves for many citizens. This implied that campaigns trying to mobilize voters on the basis of specific issue positions were often engaging with unstable preferences.

The stability of party identification — the most powerful single predictor of vote choice — created a structural anchor that campaigns operated around rather than moved substantially. The image that emerged was of party identification as a standing decision: absent compelling reason to vote differently, Democrats vote Democratic and Republicans vote Republican. The campaign's job, on this account, is to activate that standing decision among your partisans and suppress activation among the opponent's partisans.

15.1.3 What Minimal Effects Does and Doesn't Mean

The minimal effects tradition has been caricatured as claiming campaigns are irrelevant. That is not what it claims. Even within the minimal effects framework, campaigns can matter in several ways:

Activating partisans: Mobilizing your base to turn out is a form of campaign effect, even if it doesn't change preferences.

Agenda-setting: Campaigns can influence which issues voters prioritize — not changing their issue positions but changing which issues determine their vote choice.

Information effects: In low-information environments, campaigns genuinely inform voters about candidates, producing real opinion movement among the least politically engaged.

Marginal effects in close races: Even small effects — 1–2 percentage points — determine the outcome when elections are decided by small margins.

The minimal effects tradition correctly identifies that dramatic preference conversion is rare. It does not imply that campaigns are strategically irrelevant.

15.2 The Modern Revisionist Turn

15.2.1 Green, Gerber, and the Experimental Renaissance

The late 1990s and early 2000s produced a dramatic methodological shift in the study of campaign effects: the adoption of randomized field experiments as the standard for causal identification. As discussed in Chapter 14, Green and Gerber's pioneering GOTV experiments established that well-designed random assignment could produce credibly causal estimates of campaign contact effects — bypassing the endogeneity problems that plague observational studies of campaigns.

The experimental approach revealed effects that observational work had missed. Door-to-door canvassing increased turnout by 6–9 percentage points. Social pressure mailers moved turnout by 8 points. These are not trivial effects; in a competitive election, they can shift the outcome.

The experimental renaissance extended to persuasion. Studies using randomized exposure to campaign advertising, issue messages, and candidate contact produced new estimates of persuasion effects. The picture that emerged was more nuanced than simple minimal effects: some interventions, in some contexts, with some voter populations, produced meaningful preference shifts.

15.2.2 The Activation Hypothesis

One of the most important contributions of modern campaign effects research is the "activation" framework, developed most fully by John Zaller in his 1992 book The Nature and Origins of Mass Opinion and extended in his later "Floating Voters" article.

Zaller's account distinguishes between:

Distal priors: Long-standing values, partisan identities, and issue predispositions that structure political views but may not be immediately accessible.

Proximal considerations: Recently received information about candidates and issues that is more immediately salient.

Opinion formation: The integration of proximal considerations and distal priors through a sampling process — voters "report" the average of the considerations currently accessible in their memory.

Campaign communications don't primarily convert — they prime. By flooding communication channels with particular messages, campaigns can shift which considerations are most accessible at decision time, without necessarily changing underlying values. A Republican campaign that successfully elevates crime as an issue in the final weeks shifts the accessible considerations toward the Republican candidate's advantage, even among voters whose underlying crime attitudes haven't moved.

This account explains why campaigns can have substantial effects that don't show up as preference conversion in panel surveys. The voter who starts the campaign leaning Republican and ends the campaign voting Republican hasn't "converted" — but the issues that made them feel good about their vote may have been substantially shaped by campaign communication.

📊 Real-World Application: The Garza Campaign's Message Testing Nadia's analytics operation runs a weekly message-testing survey, asking a random sample of persuadable voters (those with support scores between 35 and 65) to rate their vote preference after exposure to different communication frames. The biggest mover in their August testing was an economic frame emphasizing Garza's plans for workforce development — not changing voters' economic concerns but making them feel those concerns were addressed by Garza. The issue salience, not the preference, is what the campaign can plausibly move. This is the activation hypothesis in practice.

15.3 Persuasion: What the Evidence Says

15.3.1 Persuasion Effects Are Small and Context-Dependent

The most comprehensive meta-analyses of campaign persuasion studies — including Kalla and Broockman's 2018 meta-analysis of 49 field experiments — find that average persuasion effects of campaign contact are small: roughly 0.1 percentage points per contact attempt in general elections. This is not zero, but it is much smaller than naive intuitions suggest.

Several factors modulate the effect:

Electoral environment: Persuasion effects are larger in low-information environments (primaries, local races) than in high-information environments (presidential or high-spending Senate races) where voters have already formed firm views. This matters enormously for how to interpret the evidence.

Timing: Campaign effects are larger early (before opinions have crystallized) than late (after months of exposure have settled most voters). The behavioral economics literature on primacy and recency effects in persuasion is relevant here.

Message quality: Generic issue advocacy has weak persuasion effects. Personalized, narrative-based messaging that connects to voters' existing values — the "deep canvassing" approach — shows larger effects in some studies, though the evidence remains limited.

Target population: By definition, persuasion effects are largest among genuinely persuadable voters — those who haven't made up their minds or whose partisan attachment is weak. Persuasion contact with confirmed strong partisans often produces null or backfire effects (entrenching existing views rather than shifting them).

⚠️ Common Pitfall: Confusing "Persuasion" with "Conversion" When campaigns say they are "persuading" voters, they often mean something much narrower than the textbook definition of attitude change. In practice, "persuasion" in campaign analytics usually means: contacting voters with low support scores (not already committed to your candidate) and attempting to raise those scores. This includes genuine undecideds but also weak partisans of the opposing party. The effect of this outreach is rarely to convert a committed Republican into a Garza voter; it is more often to activate a weak Republican's doubts about Whitfield or to move a true independent from Whitfield-leaning to genuinely uncertain. These are real effects, but they are not what we usually mean by attitude conversion.

15.3.2 Deep Canvassing and the Limits of Conventional Persuasion

A series of studies by David Broockman and Joshua Kalla has explored whether more intensive, relationship-based canvassing approaches can produce larger and more durable persuasion effects than conventional scripted canvassing. Their deep canvassing research — involving extended conversations that engage voters' personal experiences and encourage perspective-taking — has found meaningful attitude shifts on contentious issues including transgender rights and immigration.

The 2016 "Miami study" (Broockman and Kalla) found that a 10-minute conversation by trained canvassers produced attitude changes on transgender rights that persisted over a three-month follow-up period. This is strikingly different from the weak and ephemeral effects found for conventional canvassing on the same issues.

The practical challenge is scale. A 10-minute deeply personal conversation requires highly trained canvassers and dramatically limits the volume of contacts possible. A campaign that can run conventional canvassing at 20 contacts per canvasser-day might run deep canvassing at 6–8 contacts per day. The per-contact effect may be larger, but the total population impact depends on both effect size and volume.

🔵 Debate: Deep Canvassing vs. Broad Outreach Should the Garza campaign's limited volunteer capacity be deployed in intensive deep canvassing with a smaller number of voters, or conventional canvassing reaching a much larger universe? This is a genuine analytical question with no clean answer. If the target population includes a large number of voters who are close to the persuasion threshold (could be moved with genuine engagement), deep canvassing may win on expected-value grounds despite the volume constraint. If the target population is mostly hard partisans whose main response to outreach is mobilization (voting yes/no rather than preference change), broad conventional outreach to maximize GOTV may dominate.

15.3.3 Advertising Persuasion Effects

Television advertising has been the dominant campaign expense in American elections for decades. Does it work?

The evidence is mixed and depends heavily on what "works" means. For name recognition and basic attribute learning (voters remembering candidate positions or qualifications), advertising clearly matters, particularly early in cycles when challengers need to introduce themselves. For moving vote intention against a high-visibility incumbent in a partisan context, the effects are much weaker.

Huber and Arceneaux's 2007 paper used the geographic boundaries of broadcast markets (which don't align with state lines) to identify the causal effect of presidential advertising: voters in markets that happen to cross into closely contested states received more ads. Their design found meaningful persuasion effects of presidential advertising in low-ad environments but negligible effects in high-saturation markets where both candidates were advertising heavily — consistent with the cancellation hypothesis, where both sides' ads net out.

The most important practical implication: advertising effects decay rapidly. Studies find that the persuasion effect of a political ad has a half-life of roughly one week; after two weeks, almost nothing persists. This implies the optimal advertising strategy is to concentrate spending close to Election Day rather than running a uniform "air war" from Labor Day forward. Many campaigns defy this logic for organizational reasons (the pressure to spend early is intense), and the literature suggests they may be making a systematic error.

15.4 The Ground Game vs. the Air War

15.4.1 The Rediscovery of Field Organizing

For most of the 1990s, campaigns were in the midst of a profound shift toward television as the dominant political communication medium. Field organizing — door-to-door canvassing, phone banking, the precinct-based "ground game" — had declined in importance as television budgets grew. The consultants who managed television advertising were the kings of campaign strategy.

The Green-Gerber experimental findings triggered a major reassessment. If personal contact was eight to ten times more effective per dollar than television advertising for mobilization (and competitive on persuasion in some contexts), the rational response was to rebalance toward field. The Obama campaigns in 2008 and 2012 made exactly this rebalancing, investing heavily in neighborhood team-based organizing that deployed trained volunteers in dense peer networks rather than the traditional central-office phone bank model.

Subsequent academic and journalistic analysis of the Obama field operation estimated that the field program added 0.5–1.5 percentage points to Obama's vote share in some states — a meaningful contribution to margins that were decisive in several battlegrounds.

📊 Real-World Application: The Ground Game in the Garza Race Nadia has built a field deployment model that compares the expected vote impact of marginal dollars in television versus field. The model shows that in the state's densest media markets, television costs per persuadable impression are high and declining marginal returns are severe. In lower-density counties where field organizing is cheap (volunteers live near targets) and television is expensive relative to audience, the field ratio is favorable. The optimal strategy is not "all field" or "all TV" but a heterogeneous allocation that puts field in the places where it has comparative advantage and reserves TV for the high-efficiency markets.

Jake's perspective, shaped by experience, leans toward TV because it is controllable and scalable: you can buy air time today; you cannot train 2,000 volunteers in a week. This organizational reality is not captured in the expected-value calculation, which assumes you can implement any strategy at its theoretical optimal effectiveness. The gap between theoretical optimal and operational feasible is a recurring theme in campaign analytics.

15.4.2 Measuring Ground Game Effects

Isolating the causal effect of field organizing in actual elections is methodologically challenging. Campaigns don't randomly assign their GOTV programs; they concentrate resources in competitive areas, creating a fundamental endogeneity problem. The areas that get the most field attention are the most competitive areas, which are also the areas where external factors (national environment, advertising) are changing most rapidly.

Researchers have used several approaches to address this:

Regression discontinuity across precinct boundaries: Areas just inside versus just outside the boundary of a field program's target zone receive different amounts of field activity. Comparing outcomes across this boundary, controlling for underlying characteristics, can identify causal effects.

Natural experiments in organizational strength: Some campaigns randomize the organizational intensity of field programs (deploying more intensive organizing in a randomly selected subset of counties) for evaluation purposes. These designs are powerful but rarely available to outside researchers.

Voter-level matching: Matching voters who received field contacts with similar voters who did not, conditioning on modeled propensity to receive contact, can provide pseudo-experimental estimates though residual confounding is difficult to rule out.

The emerging consensus from these approaches is that the ground game does matter: well-implemented field organizing in competitive environments adds measurable points to vote share, above and beyond the effect on turnout alone.

15.5 Activation: Mobilizing Partisans

15.5.1 Campaigns as Activation Machines

Perhaps the most consistently supported finding in the modern campaign effects literature is that campaigns are better at activating the preferences and identities of existing partisans than at converting opponents. This "activation" effect has several components:

Information: Some voters who lean toward a party haven't yet paid enough attention to learn basic facts about the candidates. Campaign communication fills this gap, converting weak leaners into committed voters.

Enthusiasm: Partisan intensity varies across individuals and time. Campaign rallies, canvasser conversations, and effective advertising can increase emotional engagement with the election, raising the probability that weakly committed partisans follow through and vote.

Perceived stakes: Campaigns that successfully communicate that "this election matters" can activate partisans who might otherwise stay home in a low-salience environment.

The activation model implies that the universe of genuinely convertible voters — those whose vote could plausibly switch from R to D or D to R with sufficient campaign exposure — is smaller than campaigns typically assume. Much of what campaigns call "persuasion" is actually activation: moving a weak Democratic-leaning voter from "probably won't vote" to "will definitely vote for the Democrat."

15.5.2 Activation in High vs. Low Information Environments

Campaign effects through activation are largest in low-information environments, where many voters haven't yet formed clear views about candidates and issues, and smallest in high-information environments, where everyone has already been saturated with information.

Presidential elections are the highest-information environment: every voter has been exposed to years of information about the candidates, and almost no one enters October genuinely uncertain about fundamental candidate characteristics. Senate races vary enormously depending on whether the contest has national attention, candidate quality, and advertising resources.

The Garza-Whitfield race, as a competitive Senate contest in a Sun Belt swing state, occupies a middle zone. Many voters are familiar with both candidates from prior statewide exposure (Garza as attorney general, Whitfield from prior legislative races) but have not been saturated to the point of having completely fixed views. This means the campaign's communication program can still move opinion, particularly among voters who are engaged enough to receive messages but not so engaged as to have fixed commitments.

🔴 Critical Thinking: Activation as a Double-Edged Sword Activation campaigns work both ways. Garza's campaign activation may succeed in mobilizing her base. But every Whitfield rally, every attack ad, every base-mobilization email list communication is also activating Whitfield voters who might otherwise have stayed home. In a highly polarized environment where both sides are running sophisticated activation operations, the net effect on each side's relative turnout may be small. The question is not just "does our activation program work" but "does it work better than theirs" — a question that requires explicit modeling of the opposing campaign's likely activity.

15.6 Agenda-Setting: Campaigns Shape What Matters

15.6.1 The Classic Agenda-Setting Hypothesis

Walter Lippmann observed in 1922 that "the world outside and the pictures in our heads" are mediated by communication — what we think about is largely shaped by what we're told to think about, not by direct experience with the full complexity of political reality. The political science formalization of this insight came from McCombs and Shaw's 1972 study of the 1968 election, which found strong correlations between the issues emphasized by news media and the issues voters rated as most important.

Campaigns' role in agenda-setting is distinct from but related to the media's. Campaigns can attempt to elevate issues advantageous to their candidate through advertising, candidate statements, and event staging; they can also try to suppress issues disadvantageous to their candidate by starving them of campaign attention. The interaction between campaign agenda-setting and media agenda-setting is complex: campaigns that successfully insert a new issue into discourse can trigger media coverage that amplifies the effect dramatically.

15.6.2 Issue Ownership and Priming

The "issue ownership" hypothesis, developed by Petrocik and others, holds that different parties are perceived as more competent on different issues: Republicans "own" national security and crime; Democrats "own" healthcare and education. Campaigns that successfully focus the election on owned issues gain vote share, not by changing the underlying issue ownership perceptions but by changing which issues are salient at the time of decision.

Iyengar and Kinder's experimental work on agenda-setting and priming showed that manipulating the issues emphasized in news coverage altered how voters evaluated the president — priming them to weight different dimensions more heavily in their overall assessment. The parallel for campaigns is that a candidate whose advertising successfully keeps healthcare at the top of the agenda is effectively priming voters to use healthcare performance in their overall evaluation.

For Nadia, the implication is that tracking issue salience — which issues voters name as most important when surveyed — is as important as tracking candidate support scores. If Whitfield's advertising has pushed crime and border security to the top of the priority list among persuadable voters, Garza's campaign needs to either counter on those issues (risky, given Whitfield's claimed ownership) or aggressively push back to make healthcare or economic security equally salient (competing for agenda space).

📊 Real-World Application: The Agenda-Setting Battle in Competitive Senate Races Modern competitive Senate campaigns routinely run weekly "issue salience" tracking alongside horse-race tracking. The campaign war room watches not just "who are you going to vote for" but "what issues will drive your vote." A shift in the top-ranked issues from healthcare (Garza's preferred terrain) to immigration (Whitfield's) is a warning signal just as alarming as a decline in Garza's horse-race number — and it carries diagnostic information about which Whitfield communications are working.

15.7 Timing, Duration, and Decay

15.7.1 The Half-Life of Campaign Effects

One of the most practically important findings in the modern campaign effects literature is that effects decay. The effects of a campaign advertisement, a candidate visit, or a piece of direct mail are not permanent or even long-lasting. The half-life of a persuasion effect is roughly one to two weeks; the half-life of a mobilization reminder (a GOTV contact long before Election Day) may be even shorter.

Hill et al. (2013) tracked the effects of a large television advertising buy for a state gubernatorial campaign using geographic variation in ad exposure. They found that the effect of the advertising buy was substantial in the week immediately following the buy but had almost entirely decayed within two weeks. This implies that a campaign that runs its advertising in August for a November election is essentially spending money to create effects that will vanish before voters make their decisions.

Gerber et al. (2011) found a similar decay in GOTV effects: mobilization contacts long before Election Day show substantial decay, while contacts in the final two weeks before the election show much higher persistence (because the action the contact is requesting is imminent). The practical lesson is that GOTV programs should concentrate contact in the final two weeks, shifting earlier contact to persuasion or relationship-building purposes.

⚠️ Common Pitfall: Frontloading Spending Without Retention Many campaigns spend heavily in September on television advertising, then run out of money for October. The decay research suggests this is a poor strategy: September spending has largely evaporated by Election Day. The superior strategy (assuming budget constraints are fixed) is to concentrate persuasion advertising in October and GOTV contacts in the final two weeks. Campaigns that frontload spending often do so not because of evidence but because of vendor incentives (consultants are paid as a percentage of ad buys and prefer early, certain commitments) and organizational anxiety (the campaign "needs to be on the air").

15.7.2 Informational vs. Reminder Effects

The decay finding needs to be qualified by the distinction between informational and reminder effects. Early campaign communication that genuinely informs voters about candidates — who they are, what they stand for, why they should be preferred — produces a different kind of effect than late-campaign reminders to vote for someone voters already know and support.

For challengers introducing themselves to voters, early advertising is genuinely important: it creates brand awareness and candidate schema that persist and frame subsequent information. For incumbents running for reelection in high-visibility races where their record is well-known, early advertising is primarily a reminder function with short half-life.

This implies different optimal timing strategies for incumbents versus challengers. A challenger should invest early to build name recognition and a positive first impression; an incumbent should hold fire longer and concentrate spending closer to Election Day when its effects will persist to the decision point.

In the Garza race, Garza (as a non-incumbent challenging a previous senator) needed early advertising to frame herself before Whitfield's campaign defined her. Whitfield (as the Republican candidate in a right-leaning state) could afford to spend later, knowing his party brand provides a floor of support. This asymmetry shapes the competing strategies in ways that go beyond simple "who has more money."

15.8 Incumbency Advantage: Nature or Nurture?

15.8.1 The Incumbency Advantage Phenomenon

Incumbent politicians running for reelection win at dramatically higher rates than challengers in almost all democratic systems, though the effect is largest in the United States. U.S. House incumbents win reelection at rates exceeding 90 percent in most cycles. Even Senate incumbents, who face more competitive environments, win reelection at rates around 80 percent.

Several sources of incumbency advantage have been proposed:

Quality selection: The fact that incumbents won their initial election means they have been through a competitive selection process. Survivors of competitive elections may genuinely be more talented candidates than the typical challenger.

Structural resources: Incumbents have access to congressional staff, free mailing privileges (franking), government travel, and constituent service operations that function as permanent campaign activities. These are publicly funded and represent a substantial structural advantage.

Name recognition: Voters are familiar with incumbents in a way they typically are not with challengers. Name recognition reduces the information cost of voting for the incumbent and creates a status quo bias.

Position-taking: Incumbents have a record, which is a liability when the record is unpopular but an asset when the record is popular. Challengers must promise; incumbents can point to accomplishments.

GOTV infrastructure: Long-serving incumbents have accumulated donor networks, volunteer organizations, and institutional relationships that challengers must build from scratch. This organizational advantage is a form of campaign capital.

15.8.2 The Gelman-King Decomposition

Andrew Gelman and Gary King's 1990 analysis of the incumbency advantage used a sophisticated regression approach to separate the incumbency effect into components: the scare-off effect (incumbents deter quality challengers from entering the race), the experience effect (incumbents are better at campaigning), and a pure incumbency effect (the advantage of holding office per se). Their analysis suggested that all three contribute, with the scare-off effect being substantial.

The implication is that the "incumbency advantage" is partly a campaign effect (incumbents run better campaigns) and partly a structural effect (incumbents face weaker opponents). For researchers studying campaign effects, this decomposition matters: a campaign that produces a better-organized, better-funded incumbent running against a deterred-quality field looks like a structural advantage but is actually partly a campaign effect in disguise.

🔗 Connection to Later Chapters The incumbency advantage intersects with the money-in-politics literature (Chapter 36). Incumbents have dramatically easier access to campaign fundraising than challengers; they have a demonstrated ability to win, which reassures donors, and they have direct leverage over donors who want legislative access. This funding advantage is both a cause and a consequence of the incumbency advantage, creating a reinforcing cycle. Breaking into this cycle — as Garza, a non-incumbent, is attempting to do — typically requires either strong national-wave conditions or a vulnerable incumbent whose record has become a liability.

15.9 Campaign Effects in the Garza-Whitfield Race

15.9.1 When Do Campaigns Matter Most?

Political scientists have developed a rough taxonomy of when campaign effects are largest:

Close elections: In a race decided by 50.5% to 49.5%, virtually any campaign activity that moves a fraction of a percent matters. In a 60-40 race, campaigns might double or triple their effect size without changing the outcome. Competitive Senate races in swing states are where campaign effects are most consequential.

Low-information environments: Sub-Senate races — state legislative, county, and local contests — often feature candidates who are unknown to most voters. Campaign activity that builds basic name recognition and issue association can produce dramatic effects.

Open seats: Without an incumbent's structural advantage, open-seat races are more susceptible to campaign activity than incumbent-defended seats. The Garza-Whitfield contest has some characteristics of an open seat (Whitfield is not the incumbent senator, though Garza is not either).

Novel issue environments: When major new issues emerge mid-campaign — a scandal, a Supreme Court decision, an economic shock — voters who haven't yet fixed their views on the new issue are susceptible to campaign framing effects. The Garza campaign has been monitoring whether Whitfield's populist economic positions create vulnerabilities or opportunities on the economic anxiety dimension.

15.9.2 What the Structural Baseline Says

Fundamental models of election outcomes — discussed more fully in Chapter 18 — use economic conditions, presidential approval, and partisan composition to predict election results without reference to the campaigns. In a race like Garza-Whitfield, the structural baseline might predict a result within a few percentage points in either direction depending on the national environment.

If the structural baseline puts Whitfield at 52 and Garza at 48, a three-point shift is needed for Garza to win. Three points of campaign effect is substantial but within the documented range of ground game, advertising, and mobilization effects — particularly in a competitive, well-resourced Senate race.

If the structural baseline puts Whitfield at 55, Garza needs five points of campaign effect, which would be at the outer edge of the literature's documented effects. This is still possible but requires everything to go right: a highly effective ground game, successful agenda-setting on favorable issues, and significant Republican cross-over or third-party bleeding from Whitfield.

Nadia's model incorporates a structural baseline, updating it weekly with new economic indicators and presidential approval numbers. The model's uncertainty interval grows and shrinks as these structural factors move. Three weeks before Election Day, with Garza trailing by two in internal tracking, Nadia believes the structural baseline has Garza down approximately 1.5 points — meaning the campaign effects to date have been roughly neutral. She is not behind; she is roughly where the structural forces would put her without any campaign. The question is whether the final three weeks of concentrated activity can produce the marginal shift needed to win.

Jake's interpretation of the same data is grimmer: he reads the tracking number as evidence that they have spent heavily and not moved the dial. His experience tells him that campaigns that aren't winning late are usually going to lose. Both of them are right about something real.

15.9.3 The Problem of Prediction vs. Explanation

The Garza campaign's analytics operation is trying to do two different things simultaneously, and the tension between them is instructive for anyone studying campaign effects.

Prediction: What will happen on Election Day? The structural model is primarily a prediction tool. It takes inputs (economic conditions, approval, demographics) and generates a probability distribution over outcomes. It doesn't necessarily explain why those outcomes occur.

Explanation: What caused the outcome? Understanding campaign effects requires a causal model that identifies the contribution of specific activities to the outcome — not just predicting what will happen but understanding why.

These two goals are often in tension. The best prediction model might use a variable (like party registration margin in the state) that tells you a lot about likely outcomes but nothing about what campaigns can do to influence them. The best explanation model might focus on campaign activities that have genuine causal effects but are not large enough to matter much for prediction.

The most common error in campaign analytics is confusing prediction and explanation: taking the outputs of a predictive model and treating them as if they identify causal levers. If your model shows that registered Democrats have higher Garza support, that doesn't mean contacting registered Democrats will increase Garza support — most of them are already committed. Correlation (registered Democrat = Garza support) and causation (contacting registered Democrats → Garza support) are different things.

This distinction is at the heart of Chapter 15's core lesson: campaign effects are real and meaningful, but identifying them requires causal thinking that goes beyond the predictive models campaigns routinely deploy.

15.10 Synthesis: What Campaigns Can and Cannot Do

After seventy years of research, from Lazarsfeld's Erie County panel to modern field experiments, what can we say with confidence about campaign effects?

Campaigns can: - Mobilize base voters, particularly low-propensity supporters who would vote for their candidate if they vote at all - Frame issues — increase the salience of favorable issues and decrease the salience of unfavorable ones - Introduce challengers — build name recognition and positive associations for unknown candidates - Activate weak partisans — convert low-engagement partisans into committed, enthusiastic voters - Shift the composition of the electorate through differential mobilization effects - Produce modest persuasion effects among genuinely persuadable voters, particularly through intensive personal contact

Campaigns cannot: - Convert committed strong partisans of the opposing party in meaningful numbers - Overcome large structural disadvantages (a 20-point structural deficit is beyond the reach of any documented campaign effect) - Produce effects that persist without continued reinforcement — effects decay, requiring constant maintenance - Easily replicate experimental effect sizes at scale — optimal canvassing programs in controlled experiments perform better than the messy reality of a statewide campaign operation

The fundamental epistemological lesson: Campaign effects are real but difficult to identify precisely. The challenge is always separating the campaign's contribution from the structural factors that would have produced similar outcomes anyway. The best campaigns understand this uncertainty and build monitoring systems that allow real-time assessment of whether their activities are working — not just tracking their horse-race number but understanding what is driving it.

Chapter Summary

• The minimal effects tradition, rooted in Lazarsfeld's Columbia school studies, found that campaigns primarily reinforce rather than convert — selective exposure and partisan anchoring limit preference change.
• Modern revisionism, driven by randomized field experiments, has established that campaigns do have meaningful effects, particularly on mobilization, activation, and issue salience.
• Persuasion effects are real but small: the meta-analytic average is roughly 0.1 percentage points per contact in general elections, rising substantially with personal contact quality and falling in high-information environments.
• Campaigns are more effective at activating weak partisans and increasing base turnout than at converting committed opponents.
• Advertising effects decay rapidly (half-life of 1–2 weeks), implying that concentrated spending close to Election Day is more efficient than early frontloading.
• The ground game (field organizing) has larger per-dollar effects than television advertising for mobilization and is increasingly competitive for persuasion, though it is harder to scale quickly.
• Incumbency advantages are partly structural (resource access, name recognition) and partly campaign-derived (organizational capital, quality-challenger deterrence).
• In the Garza-Whitfield race, campaign effects matter because the structural baseline is competitive — a few percentage points of campaign effect can change the outcome. But separating campaign effects from structural forces requires causal thinking, not just tracking prediction models.

15.11 The Air War in the Digital Age: Social Media and Campaign Effects

15.11.1 From Broadcast to Targeted Communication

The rise of digital advertising and social media has transformed the landscape of campaign communication in ways that challenge both the minimal effects tradition and the modern experimental findings. Traditional broadcast television advertising reaches a mass audience defined by program viewership and geographic market. Digital advertising reaches algorithmically defined audiences at the individual level, enabling precision targeting that was not possible in earlier eras.

This transformation has several analytical implications. On one hand, precision targeting should improve the efficiency of persuasion efforts by concentrating messages on the genuinely persuadable rather than wasting exposures on committed partisans who won't move. On the other hand, the algorithmic media environment that makes precision targeting possible has also intensified selective exposure and filter bubble dynamics, potentially limiting the reach of cross-partisan communication.

The empirical evidence on digital advertising effects is less developed than for traditional broadcast media, partly because randomized experimental studies of digital advertising effects face methodological challenges (platforms' own targeting algorithms interact with experimental designs in ways that complicate causal inference) and partly because the technology and platforms change rapidly enough that findings from one cycle may not generalize to the next.

Facebook's internal research, published in the Journal of Political Economy by Allcott and Gentzkow and their collaborators, has found both positive and null effects of social media exposure on political knowledge and polarization, depending on the specific platform, population, and measurement approach. The large-scale Facebook experiment by Bond et al. (2012) showed that social pressure messages on voting ("Your friends have voted") produced measurable increases in turnout, but the design made it difficult to separate the platform's unique role from the social pressure mechanism that would work through any channel.

15.11.2 The Organic vs. Paid Content Distinction

Campaign effects research has focused primarily on paid campaign communication. But in the digital era, an increasingly important component of campaign communication is organic — content shared, liked, and distributed by supporters and the campaign's own followers without payment. Understanding the effects of organic content is methodologically harder because it is not randomly assigned and because the most engaged sharers are already strongly committed to one side.

The most ambitious research in this area, by Joshua Tucker and the NYU Center for Social Media and Politics, has tracked the spread of political misinformation, partisan content, and factual political information across platforms. Their general finding is that highly partisan content spreads more widely and quickly than factual content, and that committed partisans are the primary sharers. This organic amplification of partisan content compounds the selective exposure dynamic, but the causal effect on vote choice — as opposed to attitude reinforcement — remains difficult to isolate.

For campaign practitioners, the organic/paid distinction matters because it affects how you measure the reach and impact of digital communications. A viral organic post that reaches 3 million people may have very different persuasion effects than paid advertising reaching the same 3 million, because the audiences differ systematically in who they are and how they encounter the content.

📊 Real-World Application: Garza's Digital Strategy Nadia's analytics team tracks two separate metrics for the Garza campaign's digital presence: paid reach (audiences targeted by bought advertising) and organic reach (shares, likes, and views of non-paid content). They have found — consistent with the literature — that organic content performs best on base-mobilization metrics (enthusiasm, volunteer sign-ups, small donations) while paid advertising performs better on persuasion metrics (support score movement among genuinely persuadable targets). The implication for budget allocation: invest heavily in organic content to maintain base activation, and concentrate paid digital on carefully targeted persuasion universes with credible persuadability scores.

15.11.3 Viral Moments and Candidate Character Signals

Some campaign effects operate through a different mechanism than either paid advertising or field organizing: viral moments that shift voters' assessments of candidate character. A debate gaffe, a clip of a candidate saying something off-message, a news story about personal conduct — these can produce rapid shifts in support scores that neither advertising nor canvassing would generate.

The analytics challenge is separating candidate quality events (which signal something genuinely informative about the candidate) from manufactured controversies (which generate noise but may decay quickly as the story moves on). The behavioral research on candidate evaluation suggests that character assessments are among the most durable impressions voters form: first impressions of candidate competence and trustworthiness are sticky and hard to dislodge through subsequent information.

In the Garza-Whitfield race, Whitfield's populist rallies have periodically generated clips that circulate widely — sometimes helping him with his base, sometimes creating crossover resistance among moderate Republicans. Nadia's team monitors a "viral moment tracking" feed that flags rapid changes in the campaign's tracking poll that don't correspond to advertising schedule changes, attempting to identify candidate character events that are moving opinion organically.

15.12 Negative Advertising: Effects, Ethics, and the Turnout Question

15.12.1 What the Research Says About Attack Ads

Negative advertising — advertising that criticizes the opponent rather than promoting one's own candidate — has been a staple of American political campaigns for decades and a source of perennial controversy. What does the research actually say about its effectiveness?

The evidence on persuasion effects of negative advertising is mixed. Some studies find that negative ads are more memorable and more effective at changing opinions about the targeted candidate than positive ads. Others find that negative ads have higher "backlash" rates — they reduce support for the attacking candidate as well as the target, particularly when voters find the ad unfair or misleading.

The most important finding, from a broader political participation perspective, is the hotly contested "demobilization hypothesis": Ansolabehere and Iyengar's 1994 experimental research suggested that exposure to negative advertising reduced voter turnout, particularly among non-partisans, by creating a "disgust with politics" effect. If negative advertising demobilizes voters, then heavy attack campaigns could be simultaneously effective at reducing the opponent's support scores and counterproductive from a mobilization standpoint.

Subsequent research has challenged this finding. Meta-analyses find small positive, null, and small negative effects of negative advertising on turnout, depending on the study design, the population, and the type of negativity. The current scholarly consensus is that the demobilization hypothesis, while plausible, has not been robustly established — and that the most likely effect of negative advertising on turnout is near zero in most contexts.

🔵 Debate: The Normative Case Against Attack Advertising Even if negative advertising has no net demobilization effect and is effective at damaging opponents, there is a normative case against its dominance in American political communication. Advertising that focuses on opponents' failures rather than one's own affirmative vision gives voters information about reasons not to vote for someone rather than reasons to vote for anyone. A campaign environment dominated by mutual attacks may increase cynicism about both candidates without helping voters make affirmative choices. This normative concern is distinct from the empirical question of whether attack ads "work" — they can work to reduce opponent support while still being bad for democratic deliberation.

15.12.2 Contrast Ads and the Middle Path

The most empirically defensible advertising approach, based on the available research, is "contrast" advertising — advertising that directly compares the candidates on a specific dimension rather than pure attack (ignores one's own candidate) or pure promotion (ignores the opponent). Contrast ads give voters information about both candidates simultaneously, are less susceptible to fairness backlash (because the candidate is also visible and accountable), and typically perform better on both persuasion and favorability metrics than pure attacks.

In the Garza campaign, Jake favors pure contrast on healthcare — showing Garza's plan next to Whitfield's record — over pure attacks on Whitfield's personal or financial conduct. Nadia's message testing validates this preference: contrast ads on healthcare policy move persuadability scores by 1.8 points among the target population; biographical attacks on Whitfield move them by 0.9 points and have higher variance (some segments react negatively to the attacking tone). Jake's instinct about contrast is, in this instance, supported by the experimental data.

15.13 Incumbency Advantage in the Modern Era

15.13.1 The Decline of Personal Incumbency Advantage

Political scientists have documented a puzzling trend: the personal incumbency advantage in U.S. House races — the extra vote share attributable to being the individual incumbent as opposed to simply running as the incumbent party — appears to have declined in the era of high partisan polarization.

The likely mechanism is that partisan cues have become more powerful. In a highly polarized environment, voters who might have defected from their party to support a popular incumbent are less likely to do so when the partisan stakes feel higher. The "moderate incumbent" in a marginal district, who once built a personal following through constituent service and bipartisan positioning, is squeezed between a more polarized electorate that demands partisan loyalty and the strategic logic of building a broad coalition.

For non-incumbent candidates like Maria Garza, this trend has a silver lining: it means that the personal quality of her campaign may matter more relative to the structural advantage her opponent would have as an incumbent. In a highly polarized, nationalist-tide environment, every incumbent's advantage is somewhat diminished.

15.13.2 The Open Seat Dynamics

The Garza-Whitfield race has some characteristics of an open seat contest — neither candidate is a current U.S. senator from the state, and neither has the structural resource advantages of incumbency at the Senate level. What Whitfield does have is incumbency-of-identity in a state that has trended Republican: he can run as the default for Republican-leaning voters without building a complex affirmative case.

Open seat races are, as noted earlier, the environment where campaign effects are largest. Both candidates are building their electoral coalitions from a starting point that is less anchored than in an incumbent-vs-challenger contest. The ground game, the advertising, the agenda-setting — all of these have somewhat larger marginal effects because fewer voters have fully fixed views about either candidate.

This is both good news and challenging news for Nadia's analytics operation. Good news: the models' confidence intervals are wider (more uncertainty), which means there is more room for campaign activity to change outcomes. Challenging news: wider uncertainty intervals make it harder to know whether you're winning or not, and make resource allocation decisions more consequential — because the margin for error is larger when outcomes are more uncertain.

15.14 From Analytics to Decision-Making: Closing the Loop

15.14.1 The Adaptive Campaign

The most sophisticated modern campaigns don't just build models and execute plans derived from them. They build systems for rapid learning and adaptation: running small field experiments to test which messages work before scaling them, monitoring weekly tracking data for evidence of strategy effectiveness, adjusting resource allocation based on early voting patterns and real-time data.

This adaptive approach requires a specific organizational culture: one where the analytics team has genuine influence over strategic decisions, where negative evidence (the strategy isn't working) is welcomed rather than suppressed, and where the campaign is willing to change course based on data rather than doubling down on a prior commitment.

Jake's evolution over the course of the campaign illustrates what adaptive campaign management looks like in practice. He began skeptical of Nadia's propensity-score targeting and instinctively protective of base-community mobilization. By Election Day minus 14, he has updated his priors: the field experiment results from August (the relational organizing data) gave him evidence he could trust, and he's incorporated it into a modified allocation that is neither pure model-optimization nor pure gut. He's still running the campaign, but he's running it with better information and a genuine integration of the quantitative and qualitative evidence streams.

15.14.2 The Epistemological Humility Requirement

The final lesson of this chapter is epistemological: campaign effects are real but difficult to identify precisely, and the appropriate response to this difficulty is not to pretend to certainty but to manage decisions under genuine uncertainty.

Nadia knows the tracking poll has sampling error. She knows the structural model has calibration uncertainty. She knows that the field experiment results from August may not generalize perfectly to November conditions. She knows that Whitfield's campaign is doing things she can't fully observe, and those activities are affecting the environment she's trying to measure.

What she can do — and what separates good campaign analytics from bad — is: 1. Be explicit about what she knows with confidence versus what she's estimating with uncertainty 2. Use conservative assumptions when making resource allocation decisions — not optimistic ones 3. Build in monitoring mechanisms that would detect if her assumptions are wrong 4. Update her models and decisions as new information arrives, rather than anchoring too hard on prior estimates 5. Communicate uncertainty honestly to campaign leadership, rather than performing false confidence to maintain organizational trust

Jake, for his part, is better at some of these than Nadia initially gives him credit for. His skepticism of model outputs isn't irrational stubbornness — it's a reasonable prior about model reliability built from experience watching models fail. The synthesis they reach by the final weeks is better than either pure model-following or pure gut-following: it is disciplined reasoning under uncertainty, which is both the definition of good analytics and the definition of good campaign management.

Chapter Summary

• The minimal effects tradition, rooted in Lazarsfeld's Columbia school studies, found that campaigns primarily reinforce rather than convert — selective exposure and partisan anchoring limit preference change.
• Modern revisionism, driven by randomized field experiments, has established that campaigns do have meaningful effects, particularly on mobilization, activation, and issue salience.
• Persuasion effects are real but small: the meta-analytic average is roughly 0.1 percentage points per contact in general elections, rising substantially with personal contact quality and falling in high-information environments.
• Campaigns are more effective at activating weak partisans and increasing base turnout than at converting committed opponents.
• Advertising effects decay rapidly (half-life of 1–2 weeks), implying that concentrated spending close to Election Day is more efficient than early frontloading.
• Digital advertising has emerged as a major campaign expenditure with effects that are less clearly documented than broadcast television — precision targeting is a potential advantage, but selective exposure dynamics limit cross-partisan reach.
• Negative advertising's demobilization effect is empirically contested; contrast advertising (comparing both candidates) outperforms both pure promotion and pure attack in many experimental settings.
• The ground game has larger per-dollar effects than television advertising for mobilization and is increasingly competitive for persuasion, though it is harder to scale quickly.
• The personal incumbency advantage appears to have declined in the high-polarization era, but structural resource advantages of incumbency persist.
• In the Garza-Whitfield race, campaign effects matter because the structural baseline is competitive — a few percentage points of campaign effect can change the outcome. Adaptive campaign management, integrating quantitative models with organizational knowledge, produces better decisions than either pure model-following or pure intuition.
• Social media and digital advertising have reshaped the information environment but have not resolved the fundamental campaign effects questions. Precision targeting improves efficiency; selective exposure limits cross-partisan reach; and the empirical evidence on digital persuasion effects remains substantially weaker than for traditional field organizing.
• Negative advertising's demobilization effect is empirically contested; contrast advertising (comparing both candidates) typically outperforms pure attack on persuasion and favorability metrics in experimental settings.
• Epistemological humility — acknowledging and communicating uncertainty rather than performing false confidence — is both a scientific obligation and a practical requirement for effective campaign adaptation. Campaigns that confuse prediction with explanation, or model outputs with ground truth, make systematic errors in resource allocation and strategy.

A Final Note: Why Campaign Effects Research Matters Beyond Campaigns

The scientific question of whether and how campaigns affect electoral outcomes is not merely of interest to campaign professionals. It matters for democratic theory. If elections are entirely determined by structural factors — economic conditions, partisan composition, demographic trends — then campaigns are essentially irrelevant window dressing on a process whose outcome is already decided by conditions outside individual or organizational control. This is a somewhat troubling vision of democracy: the voters are moved by forces they don't control, and the campaigns that nominally inform and mobilize them have little actual effect.

The experimental evidence that campaigns do matter — that high-quality personal contact moves people, that agenda-setting changes which issues determine votes, that mobilization programs genuinely add participants — is reassuring for democratic theory. It implies that effort, organization, and resource investment actually influence outcomes. Democracy is not entirely a weather forecast that the campaigns merely observe.

At the same time, the finding that campaign effects are modest and context-dependent implies that winning elections requires both a favorable structural environment and effective campaigns. Neither is sufficient on its own. A campaign that believes it can overcome a 15-point structural disadvantage through superior mobilization is almost certainly deluded. A campaign that believes structural advantages will guarantee victory without any organizational investment is also making a mistake. The appropriate stance — which Nadia and Jake, in their best moments, embody — is to recognize the real but bounded scope of what campaigns can accomplish, and to invest accordingly.

Next: Chapter 16 puts these analytical concepts into practice, using Python to visualize the electoral geography of a state like the one in which Garza and Whitfield are competing.