Information and voting: Evidence from Peru's 2026 general elections

arXiv:2606.01687 · econ.EM · Submitted 2026-06-01 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.

Ruth: I'm Ruth.

Warren: And I'm Warren, here with Iris.

Ruth: We're looking at the paper: "Information and voting".

Warren: Here's the gist of the paper. We study how election-night flash estimates shape voting in Peru’s fragmented 2026 presidential election This paper is a Bayesian-updating model of multi-candidate plurality voting frames the analysis, yielding predictions about vote reallocation toward the three candidates the estimates rendered viable—López Aliaga, Sánchez, and Nieto The Natural Experiment The study exploits a natural experiment where on 12 April 2026, 187 polling tables across 13 voting centers failed to install, and the Jurado Nacional de Elecciones (JNE) extended voting for the affected ≈ 55000 electors to Monday 13 April These tomorrow voters cast their ballots after observing the Ipsos and Datum flash estimates; otherwise comparable Sunday voters did not.

Ruth: The authors identify the treatment as a change of Blackwell (1953) experiment over the state space Θ of Section 4.2.

Warren: The Information Design Model The model develops a tractable formulation in which each voter’s choice probability is a multinomial logit in which the posterior viability of a candidate is interacted with the voter’s expressive match to him.

Ruth: The instrumental pull of viability is scaled by expressive affinity, so a nationally viable but expressively distant candidate gains where his support is concentrated and stays flat elsewhere.

Title and authors: Ruth: Let's talk about who wrote this and what this paper is actually aiming at, starting with the title "Information and voting: Evidence from Peru's two thousand twenty-six general elections <ref:2606.01687#pg1,Information and voting: Evidence from Peru's 2026>." It immediately tells you the subject matter is how external information impacts electoral outcomes in a specific, volatile place.

Warren: And the authors are Marcelo Gallardo, Nicolas Velarde, and Cristina Gutarra. They’re bringing an economist’s view on markets and incentives to this kind of political system.

Iris: From an econometrician's perspective, the title signals that they are testing a causal link: information causing a change in behavior. It suggests they are moving beyond just correlation to see if the flash estimates actually *do* something measurable.

Ruth: That’s right. What’s interesting is that they specify two thousand twenty-six and Peru, which grounds the abstract idea of information design into a very concrete political reality for anyone tracking Latin American politics right now <ref:2606.01687#pg1>.

Warren: And what this work is aiming to do is show exactly how those late signals—the Ipsos and Datum estimates—reallocate votes toward three specific candidates they rendered viable: López Aliaga, Sánchez, and Nieto.

The paper's summary: Ruth: So the summary explains that the study uses a Bayesian-updating model to frame this analysis. It’s not just saying "the estimates matter," it’s modeling precisely how the perceived viability of a candidate interacts with how much that voter actually likes him.

Warren: They introduce something called an information design model where each voter's choice probability is a multinomial logit, and they link the posterior viability of a candidate to the voter’s expressive match to him.

Iris: The key mechanism they highlight is scaling that instrumental pull of viability by expressive affinity. This means a nationally viable candidate gains support where their local preference is strongest, but stays flat elsewhere.

Ruth: That’s a big point because it moves beyond simple viability updates; it shows that the reallocation isn't just about who's leading nationally, but where people are already leaning locally.

Warren: They define a function called the perceived-viability function, qc, which maps the posterior belief about runoff entrants to the actual probability that a candidate is in that runoff.

The paper's improvements: Ruth: Now for what they suggest improving—the authors point out that their model intentionally abstracts away other information channels like rapid counts or social media commentary. They acknowledge this is a deliberate restriction for now, but it points toward future work.

Warren: And the paper suggests incorporating competing public signals, like having conflicting flash estimates from different polling firms, to see how those disagreements affect vote reallocation based on both viability and expressive match.

Iris: I think that’s a very smart suggestion because in a real election, you often get conflicting signals, and understanding how voters process those conflicts is key to predicting shifts.

Ruth: It moves the system beyond just one signal—one flash estimate—to account for the noise in the information environment itself. It shows that moving forward, we need models that handle competition between different sources of viability data.

Conclusion: Ruth: To wrap up, this paper on "Information and voting: Evidence from Peru's two thousand twenty-six general elections" shows that the flash estimates did indeed raise the perceived viability for López Aliaga, Sánchez, and Nieto <ref:2606.01687#pg1,Information and voting: Evidence from Peru's 2026>. The basic prediction was that those viable candidates would gain support while non-viable ones would lose it.

Warren: But the results are more nuanced than just a simple gain or loss; they show that the reallocation is overwhelmingly concentrated on Jorge Nieto, who gained seventeen point three percent because he had the highest match with local voters.

Iris: That ordering—Nieto greater than López Aliaga greater than Sánchez—is what’s really telling us about how local ideological affinity dictates the vote flow, not just which candidate looks best nationally.

Ruth: Precisely. The finding is that a candidate can get a national signal, but their local gain is governed by the match to voters, and Nieto’s strong match made him take the bulk of those new votes.

Warren: So for someone listening to this, it means that when you see late information about who's viable in an election like this, you need to think about how much that candidate actually aligns with your local viewpoint to predict their actual vote share change.

Marcelo Gallardo, Nicolas Velarde, Cristina Gutarra

Pontificia Universidad Catolica del Peru (PUCP) · University of California, Berkeley

econ.EM

Submitted: 2026-06-01

Updated: 2026-10-31

Comments: Accepted for presentation at the LACEA-LAMES 2026 Annual Meeting

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 91/100

The gist: We study how election-night flash estimates shape voting in Peru’s fragmented 2026 presidential election This paper is a Bayesian-updating model of multi-candidate plurality voting frames the…

Key concepts

Natural Experiment
The study used a real-world event: polling tables failed to install in some areas, delaying voting for about 55,000 people until Monday. This created a situation where some voters saw the flash estimates before others, allowing researchers to compare those who saw the estimates versus those who didn't.
Information Design Model
This model explains how a voter chooses between candidates based on two factors: whether a candidate is deemed 'viable' (based on polling) and how much that candidate matches the voter's personal political views. Viability pulls support, but affinity scales how much support actually moves.
Expressive Affinity
This refers to a voter's inherent ideological preference or match with a candidate. The key finding is that while viability attracts votes, the final reallocation is governed by this affinity; candidates who are ideologically closer to the voters gain more local support.
Perceived Viability Function (qc)
This function measures how likely a voter believes a specific candidate will actually enter the runoff. The flash estimates directly increase this perceived viability score for three candidates, signaling their potential to gain votes.

Terminology

Summary

We study how election-night flash estimates shape voting in Peru’s fragmented 2026 presidential election This paper is a Bayesian-updating model of multi-candidate plurality voting frames the analysis, yielding predictions about vote reallocation toward the three candidates the estimates rendered viable—López Aliaga, Sánchez, and Nieto

The Natural Experiment

The study exploits a natural experiment where on 12 April 2026, 187 polling tables across 13 voting centers failed to install, and the Jurado Nacional de Elecciones (JNE) extended voting for the affected ≈ 55000 electors to Monday 13 April These tomorrow voters cast their ballots after observing the Ipsos and Datum flash estimates; otherwise comparable Sunday voters did not. The authors identify the treatment as a change of Blackwell (1953) experiment over the state space Θ of Section 4. p. 82.

The Information Design Model

The model develops a tractable formulation in which each voter’s choice probability is a multinomial logit in which the posterior viability of a candidate is interacted with the voter’s expressive match to him. The instrumental pull of viability is scaled by expressive affinity, so a nationally viable but expressively distant candidate gains where his support is concentrated and stays flat elsewhere. p. 1 The perceived-viability function qc maps a posterior belief over runoff entrants into the perceived probability that c is in the runoff. p. 1

The Empirical Strategy

The empirical section exploits the JNE ruling to identify the treated locales de votación, builds a new socioeconomic dataset from INEI Redatam census data and 2021 results, and estimates the average treatment effect on the treated by cardinality matching (Zubizarreta, Paredes, and Rosenbaum, 2014). The estimand is the acta-weighted average of τac across the treated population. p. 0

Key Findings

The flash raises perceived viability qc(µs j) above the prior for three candidates—Nieto (within a percentage point of both named candidates in either estimate), López Aliaga (Datum’s named candidate), and Sánchez (Ipsos’s named candidate). The basic prediction is that the viable candidates gain and the non-viable candidates lose. p. 0 The reallocation is overwhelmingly concentrated on Nieto, whose +17. p. 13-point gain is three to seven times the gain of any other candidate. p. 5 The ordering Nieto > López Aliaga > Sánchez tracks affinity, not national viability. p. 1

Robustness and Limitations

The headline pattern survives every exercise of the sensitivity battery of Section 6.8, including leave-one-treated-location-out and Rosenbaum bounds. p. 1 The main limitation is the small number of treated locations, as the independent causal information comes from LT = 13 clusters. p. 5 The design relies on matching only observed covariates, so sensitivity to unobserved confounding remains a concern. p. 5

Conclusion

The contribution is to identify the causal effect of a late viability signal on vote reallocation and to show that it runs through the local distribution of expressive affinities, not the signal alone. Information has a significant, robust effect. p. 4 The flash raised the shares of the three candidates it rendered viable (Nieto, López Aliaga, Sánchez; H1–H3), weakly reinforced Fujimori (+2. p. 06 points; H4), and lowered the non-viable candidates’ shares (H5). p. 0 Every effect is significant at 1% and survives the leave-one-out, weighting, tolerance, and Rosenbaum checks. p. 1 The magnitudes follow Mj(c): because qc moves uniformly while the reallocation is scaled by match, the ordering Nieto > López Aliaga > Sánchez tracks affinity, not national viability. p. 1 Nieto, the off-axis “clean” option with the highest match, takes the bulk (+17. p. 63 points); López Aliaga’s match is depressed by his tie to the congressional pact (+5. p. 61); Sánchez, the eventual runoff entrant, is expressively most distant in this right-leaning sample and gains least (+2. p. 17). A candidate can thus advance nationally on a signal yet move few local votes, because viability travels with the signal while its reallocation is governed by local affinity. p. 1

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Multi-candidate plurality vote prediction systems can be improved by incorporating information design models that explicitly account for two competing public signals, such as conflicting flash estimates from different polling firms, to predict vote reallocation based on viability and expressive match. This allows the system to move beyond simple viability updates to capture how the informational shock raised the vote share of the candidates that the flash estimates rendered viable while respecting local ideological affinities.

  2. Causal inference frameworks can be enhanced by implementing matching procedures that explicitly control for complex, multi-dimensional covariate imbalances using cardinality matching based on absolute standardized mean differences (SMD). This ensures that treatment effects are identified via cardinality matching (Zubizarreta, Paredes, and Rosenbaum, 2014) rather than relying solely on propensity score models.

  3. The AI system can be trained to predict candidate-level vote share changes in specific local areas by utilizing a behavioral utility function that includes an interaction term between perceived viability and expressive match. This allows the system to capture the finding that a rise in qc(mu s j) raises Pj (c s) by an amount increasing in Mj (c), meaning reallocation is not just toward viable candidates but is scaled by local ideological acceptability.

  4. The system can be designed to distinguish between national viability gains and local gain by analyzing the signs of the estimated treatment effect across different candidate profiles. This enables the AI to predict that "the ordering Nieto > López Aliaga > Sánchez tracks affinity, not national viability," providing nuanced predictions on which candidates gain votes based on their specific ideological positioning relative to Lima Sur voters.

  5. The system can incorporate robustness checks against unobserved confounding by calculating Rosenbaum bounds for treatment effects. This allows the AI to quantify the worst-case p-value associated with each electoral outcome as the sensitivity parameter Gamma increases, providing confidence metrics on how robust the predicted vote reallocation is to hidden selection factors.

Related papers