Introduction
Inflation re-emerged as a major global economic challenge following the Covid-19 pandemic. In advanced industrialized democracies, the rate of inflation increased to levels not seen in forty years (Ha et al. Reference Ha, Kose and Ohnsorge2023). Fast-rising prices are widely believed to reduce the popularity of incumbent governments, and post-pandemic elections in countries with high inflation provide some anecdotal support for this expectation. For instance, incumbent parties experienced major declines in their vote shares in Argentina, India, the United Kingdom, and the United States – four countries where inflation had been high by historical standards.Footnote 1
This study examines the causal effect of inflation on support for incumbents. We test the effect of inflation in one crucial case: the 2024 US Presidential election. In this election, the Republican challenger, Donald Trump, made inflation a centerpiece of his campaign. At a rally in New York’s Madison Square Garden just days before the election, the Republican nominee presented his closing argument, first asking the crowd if they were ‘better off now’ than four years ago, then adding: ‘With your vote in this election, I will end inflation’.Footnote 2 In the days following the election, many journalists and pundits singled out inflation as a critical factor behind Trump’s decisive victory (Burns Reference Burns2024; Delaney Reference Delaney2024; Elliott Reference Elliott2024; Lowrey Reference Lowrey2024; McArdle Reference McArdle2024).
Previous scholarship suggests that these conclusions are plausible. Numerous studies have found that incumbent vote shares are negatively correlated with inflation (Chappell and Veiga Reference Chappell and Veiga2000; Fair Reference Fair1996; Goodman and Kramer Reference Goodman and Kramer1975; Kiewiet and Udell Reference Kiewiet and Udell1998; Lewis-Beck Reference Lewis-Beck1990; Palmer and Whitten Reference Palmer and Whitten1999). Time-series studies likewise find that incumbent approval ratings decline when inflation is high (Beck Reference Beck1991; Berlemann and Enkelmann Reference Berlemann and Enkelmann2014; Carlin et al. Reference Carlin, Hartlyn, Hellwig, Love, Martínez-Gallardo and Singer2018). Research based on individual-level survey data also finds a link between inflation and anti-incumbent attitudes. People with negative perceptions of inflation tend to express lower levels of government approval and are less likely to vote for the incumbent party (Baccini and Weymouth Reference Baccini and Weymouth2025; Clarke and Whiteley Reference Clarke and Whiteley1990; Lewis-Beck and Stegmaier Reference Lewis-Beck and Stegmaier2000; Price and Sanders Reference Price and Sanders1995).
On the other hand, there are reasons to question whether inflation influences incumbent evaluations in contemporary elections, especially given the high levels of partisan polarization observed in many countries (Boxell et al. Reference Boxell, Gentzkow and Shapiro2024; Reiljan et al. Reference Reiljan, Garzia, Ferreira da Silva and Trechsel2024). Studies have shown that partisan polarization attenuates the link between economic conditions and incumbent support (see, for example, Donovan et al. Reference Donovan, Kellstedt, Key and Lebo2020; Ellis and Ura Reference Ellis and Ura2021; Moreira Reference Moreira2025). Drawing on county-level data from US Presidential elections between 1992 and 2020 and ANES surveys during the same period, Moreira (Reference Moreira2025) shows that as polarization rises, voters become less likely to punish or reward the incumbent based on economic conditions or perceptions. Consistent with this observation, Mutz and Mansfield (Reference Mutz and Mansfield2024) find that individuals who had worse views of inflation were no more likely to switch their votes away from the incumbent Democrats in the 2022 US Congressional elections. Another study of that election finds that perceptions of inflation influenced vote choices in the 2022 Congressional elections, but objective indicators of individuals’ exposure to inflation did not (Baccini and Weymouth Reference Baccini and Weymouth2025).Footnote 3
Moreover, existing studies on the relationship between inflation and incumbent approval face a potential endogeneity problem. Economic perceptions, such as inflation assessments, are influenced by individuals’ political leanings (see, for example, Evans and Andersen Reference Evans and Andersen2006; Evans and Pickup Reference Evans and Pickup2010). Research shows that partisanship has a sizable impact on individuals’ inflation expectations, both in the United States and elsewhere (Bachmann et al. Reference Bachmann, Gründler, Potrafke and Seiberlich2021; Binder et al. Reference Binder, Kamdar and Ryngaert2024; Copelovitch and Wagner Reference Copelovitch and Wagner2024; Gillitzer et al. Reference Gillitzer, Prasad and Robinson2021; Stantcheva Reference Stantcheva2024). Consequently, prior studies that establish a correlation between individuals’ inflation perceptions and their support for incumbents are unable to support a causal interpretation of this relationship.Footnote 4
To overcome the limitations of previous studies, we adopt an experimental design.Footnote 5 Our pre-registered experiment was embedded in a large, nationally representative, survey fielded the week prior to the 2024 US Presidential election. Specifically, we employed a question-order design in which some respondents were asked to report about how much prices have gone up in the last year before being asked for their evaluations of the Biden–Harris administration and the Democratic Party. Other respondents received the inflation question after providing those evaluations. This design enables us to test whether thinking about inflation has a causal effect on support for the incumbent party.
Our results paint a clear picture that inflation reduced voter support for the Democratic Party in the 2024 Presidential election. Respondents who were first primed to think about inflation reported lower approval of the Biden–Harris administration and expressed less confidence in the ability of the Democratic Party leadership to manage the economy. This drop in approval is particularly steep among two critical groups that Harris needed to turn out to support her at the election: voters that identify as Democrats and those that identify as political Independents. These patterns provide new evidence that the salience of inflation contributed to the Democratic Party’s poor performance in the 2024 Presidential election. More broadly, this study provides new, causally identified, evidence that inflation can reduce incumbent popularity.
While we find that the economy still influences incumbent approval even in a context of high political polarization, some voters’ assessments of the incumbent appear to be divorced from the state of the economy. Specifically, our inflation prime had no effect on approval ratings among Republican identifiers. This suggests that in polarized contexts, some individuals may have such intense dislike of the incumbent party that their evaluations of the incumbent are unaffected by the level of inflation or other economic conditions. Thus, while our study implies that polarization does not eliminate the influence of inflation on incumbent popularity, it may reduce its influence for a segment of the voting public.
Research Design and Hypothesis
Our experiment was included in a large, nationally representative survey fielded in the United States just days before the 2024 presidential election.Footnote 6 The experiment manipulated the order in which respondents received some survey questions. Respondents in the treatment group received a question about the inflation rate immediately before being asked for their evaluations of the incumbent Democratic Party. Those in the control group received the inflation question after incumbent evaluations.Footnote 7 Receiving the inflation question first should prime respondents to focus on this topic when answering questions about the incumbent.
The question about inflation asked subjects how much prices for goods and services had changed in the last year. We presented five answer categories, ranging from prices being lower than they were a year ago to prices being more than 20 per cent higher than a year ago. The exact wording of the question-and-answer categories read as follows:
Think for a moment about how much you pay for goods and services today, and how much cheaper or more expensive everything has become over the past year. In your experience, which of the following best describes what has happened to the prices that you pay for things compared to one year ago?
(1) Prices today are lower than they were one year ago; (2) Prices today are about the same as they were one year ago; (3) Prices today are 1–5% higher than one year ago; (4) Prices today are 6–20% higher than one year ago; (5) Prices today are more than 20% higher than a year ago.
It is worth emphasizing that we did not provide any positive or negative information about inflation to respondents. Instead, our design requires individuals to rely on their pre-existing knowledge and understanding of inflation when formulating their political attitudes. Thus, our empirical strategy utilizes an indirect treatment (Mutz Reference Mutz2011, 50) designed to ‘surreptitiously “prime” or bring to mind a particular consideration’ without the respondent being aware of the relevant characteristic of the treatment. The subtlety, or low ‘dosage’, of our treatment limits the risk that our design will produce unrealistically strong treatment effects (Barabas and Jerit Reference Barabas and Jerit2010; Brutger et al. Reference Brutger, Kertzer, Renshon and Weiss2022). Moreover, the effect of this treatment should closely mimic the effects of increased attention to inflation that occurs in the real world when consumers experience higher prices at the grocery store and gas station, or when the press or opposition parties highlight this issue. The close correspondence between our treatment and how people receive this information in their everyday lives should enhance the external validity of our study (Findley et al. Reference Findley, Kikuta and Denly2021).
Two outcomes of interest serve as our dependent variables. First, we asked respondents the extent to which they approved or disapproved of the Biden–Harris administration’s performance on a scale from 0 (‘strongly disapprove’) to 10 (‘strongly approve’). Second, we asked about respondents’ level of confidence in the ability of the Democratic Party leadership to manage the economy, again on a 0–10 scale (0 = no confidence, 10 = extremely confident). This way we can assess respondents’ evaluations of the incumbent government broadly, as well as their evaluations of the Democrats’ competence managing the economy specifically.
We expect receiving the inflation question prior to the government evaluation questions should influence responses to the latter questions through a process known as ‘priming’. Research on priming finds that issues that receive more attention from the media and political campaigns become more salient in voters’ minds and have a disproportionately large impact on how voters evaluate political candidates (Edwards et al. Reference Edwards, Mitchell and Welch1995; Hart Reference Hart2013; Iyengar and Kinder Reference Iyengar and Kinder1987; Krosnick and Kinder Reference Krosnick and Kinder1990; Vavreck Reference Vavreck2009). Within the context of a survey, the topic of the preceding question should, similarly, increase the salience of that topic.Footnote 8 Thus, compared to those in the control group, voters that receive the inflation question before the outcome questions are likely to focus more on inflation when formulating their opinions about the Democratic Party.
We anticipate that priming respondents to think about inflation will reduce average levels of incumbent approval in general and confidence in the incumbent’s management of the economy more specifically.Footnote 9 Priming individuals about inflation is likely to influence incumbent evaluations through both pocketbook and sociotropic channels. Since Americans view inflation as an unambiguously negative outcome (Stantcheva Reference Stantcheva2024), heightening the salience of inflation may lead voters to focus on their declining purchasing power, and this perceived decline in their personal pocketbook situation is likely to worsen their evaluations of the incumbent party. Additionally, primes about inflation may operate through sociotropic channels, such as by creating a perception that the incumbent party lacks competence. For example, a majority of American voters believed that the Republican candidate, Donald Trump, would handle inflation better than Kamala Harris, the Democratic Party’s nominee for President (Ballard Reference Ballard2024). Getting voters to focus on an issue in which the incumbent party had a poor track record may therefore heighten the perception that the incumbent lacks competence, thereby reducing their popularity. For these reasons, we expect the following:
Hypothesis 1: Priming individuals to think about inflation reduces incumbent approval and confidence in the incumbent party’s ability to manage the economy.
Findings
Inflation Perceptions and Incumbent Evaluations
Before turning to our main experimental findings, we first explore whether reported inflation perceptions correlate with our two outcome questions: approval of the Biden–Harris administration and confidence in the Democrats’ management of the economy. In line with previous observational studies, Appendix B shows that individuals who report higher levels of inflation are more likely to report negative views of the incumbent party. However, we also find that party identity is a very strong predictor of inflation perceptions; Republican identifiers report significantly higher inflation rates than Democrats. Thus, rather than inflation causing low approval of the incumbent party, it may be the case that low approval of the incumbent party causes individuals to report higher price increases. We now turn to our question-order experiment, which was designed to overcome these causal-identification challenges.
Main Experimental Results
Our main findings are straightforward: priming voters to think about inflation reduces support for the incumbent party. The average level of approval of the Biden–Harris administration among those who were not first primed to think about inflation (the control condition) is 4.7. On the other hand, respondents who were first asked to report price increases (the treatment condition) return a mean of 4.2, a difference of 0.5 points. This difference is statistically significant (p < 0.01). Average scores for the question about confidence in the Democratic Party leadership’s management of the economy are 4.8 in the control condition and 4.5 in the treatment condition, and this difference is also statistically significant (p < 0.05).
Figure 1 plots our estimated average treatment effects, along with 95% confidence intervals. Following our pre-analysis plan, we present two sets of estimates: those without any control variables and those from an ordinary least squares (OLS) regression model that includes controls for individuals’ party identification, level of education, gender identity, age, and state fixed effects.Footnote 10 The inclusion of controls produces slightly smaller and less precise point estimates than those reported above.Footnote 11 Nevertheless, the evidence is largely consistent with the hypothesis that raising the salience of inflation leads to worse evaluations of incumbent parties and their candidates.Footnote 12
Figure 1. Average treatment effects.
Note: Horizontal lines display 95 per cent confidence intervals around the estimates.
The sizes of these effects are substantively meaningful. By way of comparison, in the model with control variables, the difference in average approval rates between the treatment and control groups is about as large as the estimated difference in approval rates between those with bachelor’s degrees and high school graduates. The estimated effect of our treatment in that model is also 60 per cent larger than the gap in approval rates between men and women. To further illustrate the substantive effects of the treatment, Figures C1 and C2 in the Appendix present the full distribution of responses to both outcome questions across the two experimental conditions. Those figures show that the treatment leads to a noticeable decline in the percentage of respondents that express the strongest rates of approval and highest levels of confidence in the Democratic Party. The treatment reduces the percentage of respondents that express the highest rates of approval and confidence by 30 per cent and 23 per cent, respectively.
Although our sample is broadly representative of the US population, it skews slightly female, more educated, and older than the overall population. To assess whether this impacts our results, as a robustness check we also applied post-stratification weights based on demographic variables of gender, age, and level of education. The results, presented in Table C2, are consistent with our unweighted estimates. In fact, the estimated effects with sampling weights are larger, and are statistically significant at the 95 per cent confidence level in all specifications.
Heterogeneity across Partisan Groups
Next, we examine whether inflation has different effects on different groups of voters. We focus our attention on partisanship because individuals’ party identities have very strong effects on their vote choice and government evaluations.Footnote 13 While it seems plausible that partisanship would condition voters’ responses to inflation, it is less obvious whether Republicans, Democrats, or Independents should be most responsive to our treatment. One possibility is that Republicans, because they report stronger concern with inflation than Democrats in recent US surveys (Binetti et al. Reference Binetti, Nuzzi and Stantcheva2024; Stantcheva Reference Stantcheva2024), should be most affected by the inflation prime. On the other hand, because most Republicans oppose the Democratic Party’s policy positions on many issues, Republicans may oppose the incumbent irrespective of the inflation rate due to ideological differences. Hence, Democratic identifiers, because they are more sympathetic ideologically with the incumbent, may be more susceptible to primes about inflation. A third, and final, possibility is that the treatment may have the strongest effect among political Independents because they lack strong partisan attachments and might be particularly inclined to assess candidates based on their performance.
To examine whether and how partisanship moderates the effect of the inflation prime, we add interaction terms between the treatment and individuals’ party identification to the multiple regression models estimated above.Footnote 14 Figure 2 plots the conditional treatment effects for the two outcome variables across three partisan groups: those that identify as Democrats, Independents, and Republicans.Footnote 15 We find that partisanship moderates the effect of the inflation prime. Among Democrats, the treatment reduces approval by 0.5 and lowers confidence by 0.3 points, and both effects are statistically significant. We obtain similar point estimates for Independents, though the effect on confidence is only statistically significant at the 90 per cent confidence interval, presumably due to the smaller number of Independents in our sample. Republicans, by contrast, were unmoved by the treatment.Footnote 16 The difference in effects between Republicans and both Democrats and Independents is statistically significant.
Figure 2. Heterogeneous treatment effects: partisanship.
Note: Horizontal lines display 95 per cent confidence intervals around the estimates.
This weak effect among Republicans is likely due to their very strong pre-existing disapproval of the Biden–Harris administration. For instance, 63 per cent of Republican identifiers in the control condition reported the strongest level of disapproval, leaving little room for the inflation prime to further lower this support (see Figure D1 in the Appendix). By contrast, only 22 per cent of Independents in the control condition strongly disapproved of the Biden–Harris administration, and the treatment increased this proportion to 28 per cent. Meanwhile, the modal Democrat in the control condition reported the strongest level of approval for the Biden–Harris administration. The treatment lowered the proportion of Democrats in the top category from 25 per cent to 17 per cent. In sum, raising the salience of inflation hurt the Democrats in two ways: it increased the number of Independents that had very negative opinions of the party and lowered the proportion of Democrats who were highly enthusiastic about their own party.
Heterogeneity across Individuals’ Exposure to Inflation
Lastly, we explore whether the treatment has a stronger effect on individuals that have personally experienced higher rates of inflation. We consider two measures of exposure to inflation. The first is the subjective measure based on responses to the inflation-perceptions question in the survey. Second, we utilize a more objective indicator of inflation exposure: the cumulative rate of inflation over the previous four years (2021–2024) within the Census division in which that individual resides.Footnote 17
To test whether inflation exposure moderates the effect of the treatment, we add these measures of inflation experience as well as interaction terms with treatment status to the baseline regression model (controlling for the same set of pretreatment covariates as before). Individuals’ inflation perceptions do not moderate the effect of the treatment (see Table E1). As discussed earlier, responses to the inflation-perception question are very partisan in nature. This null result may therefore be explained by the lack of sincere responses to this survey question.
However, we do find that the treatment has a larger effect, both substantively and statistically, in parts of the country where inflation went up more rapidly based on objective data. Figure 3 plots the estimated effect of the treatment at different values of regional inflation (see also Table E2). It shows that the treatment has small and statistically insignificant effects in relatively low-inflation regions. We estimate treatment effects of –0.08 and 0.01 across the two dependent variables in a region such as the Middle Atlantic where the price level increased about 19 per cent in this period. By contrast, the treatment has large negative and statistically significant effects in regions where inflation was relatively high. For instance, in the South Atlantic, where prices went up over 22 per cent in this period, we estimate that the inflation prime reduced approval of the Biden–Harris administration by 0.56 and reduced confidence in the Democratic Party’s leadership by 0.34 points. As we also show in Table E2, this pattern holds even when controlling for the interaction between the treatment and individuals’ partisan identities, which suggests that these results are not simply picking up the effects of partisanship.Footnote 18 These results suggest that priming individuals about inflation may not work in all circumstances but is particularly likely to turn voters against the incumbent party when inflation rates are higher.
Figure 3. Heterogeneous treatment effects: regional inflation rates.
Note: Solid lines display marginal effects; dashed lines display 95 per cent confidence intervals.
Discussion and Conclusions
Prior scholarship has clearly established that inflation is unpopular with voters (see, for example, Stantcheva Reference Stantcheva2024). However, most of the evidence linking it to electoral outcomes is far from definitive because previous studies are unable to rule out the possibility that inflation perceptions are reflections of political evaluations rather than the cause of those evaluations. To ameliorate this challenge, this study relies on an experimental design – specifically, a question-order survey experiment that subtly primes some respondents to self-report on price increases prior to offering their evaluations of the incumbent government.
We find that priming voters on inflation significantly and substantively reduces approval of the Biden–Harris administration and lowers confidence in the Democratic Party leadership’s ability to manage the economy. When we condition our results on partisan identity, we get an even more granular understanding of how inflation shaped the views of the 2024 electorate. Interestingly, our treatment does not diminish Republican voters’ evaluations of the government, likely reflecting the fact that this group already thought poorly of the Biden–Harris administration. On the other hand, we find that the inflation treatment effect is concentrated among Independents and Democratic voters. This finding is notable given that a Harris victory largely hinged on whether she could sway Independents to her side while also turning out Democrats in large numbers. In light of these results, it is reasonable to conclude that inflation played a critical role in the Harris–Walz defeat by tarnishing the ticket’s reputation among Independent voters and dampening enthusiasm among the Democratic Party faithful. We also find that priming people about inflation only reduced incumbent approval in regions of the United States where prices had gone up particularly rapidly during the four years of the Biden presidency.
Our results are therefore highly suggestive that inflation impacted the outcome of this election, but several caveats are in order. The ability of survey experiments to generalize to real world settings can be compromised if the treatments and outcomes ‘do not resemble the relevant phenomena in question or if the experimental setting exaggerates the effect of the stimulus’ (Barabas and Jerit Reference Barabas and Jerit2010, 226). The outcome variables in our study do not measure voting directly; instead, we examine measures of support for the incumbent, which are strongly correlated with voting intentions.Footnote 19 Similarly, our treatment cannot manipulate individuals’ actual inflation experiences. However, using an unobtrusive prime as our experimental manipulation, which does not provide respondents with any new information or frames, should improve the external validity of our findings. By simply asking some respondents to think about inflation, our treatment approximates how individuals receive information about inflation in everyday settings, such as observing higher prices in grocery stores and consuming news articles about inflation. In short, while our results cannot establish that inflation impacted how individuals voted in this election, they demonstrate that raising the salience of inflation reduced popular support for the Democratic Party during the 2024 US Presidential election period.
More broadly, our study provides new, causally identified evidence that macroeconomic conditions in general, and inflation in particular, can influence incumbent popularity. Our evidence also suggests that the economy remains influential even when partisan polarization is high. At the same time, however, the lack of treatment effects among Republican voters does suggest that political polarization may reduce the importance of the economy for some types of voters. Our findings also do not imply that inflation will matter in all elections. This may depend on, among other factors, the level and duration of inflation, as well as the ability of voters to attribute responsibility for inflation to incumbents – a factor that could potentially vary across executive and legislative elections and across different electoral systems. Given ongoing concerns about inflation in the post-Covid era, it would be useful for researchers to examine whether inflation has similar effects on incumbent approval, and among which groups, in other national contexts.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0007123425100811
Data availability statement
Replication data for this paper can be found in Harvard Dataverse at https://doi.org/10.7910/DVN/VLCCUV.
Acknowledgments
For their constructive feedback on earlier drafts of this article, we thank Toby Bolsen, Susan Stokes, Stephen Weymouth, and the editor and anonymous reviewers at BJPS.
Author contributions
Authors’ names are listed in alphabetical order.
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interests
None to disclose.
Open access
