Industrial Policy Projects Boosted Harris and Hurt Trump in the 2024 Election, but Not by Much

The Biden-Harris administration’s signature industrial policy program—supporting investment in new factories, roads, bridges, and other infrastructure and manufacturing projects across the country through government grants, loans, and tax breaks with the Infrastructure Investment and Jobs Act (IIJA), the CHIPS and Science Act (CHIPS), and the Inflation Reduction Act (IRA)—tested the theory that presidential candidates can reap electoral rewards by delivering good jobs and meaningful economic opportunity for the working class.¹

This report finds that federal investments in local industrial policy projects were associated with an improvement in county-level vote share for Kamala Harris in 2024 compared with Joe Biden’s county-level vote share in 2020. Furthermore, projects were associated with a decrease in Donald Trump’s county-level vote share in 2024 compared with 2020. However, the vote share differences were slight and far from sufficient to swing the election in Harris’ favor. In particular, the Center for American Progress analysis of county-level data finds:

1. Each additional IIJA, CHIPS, or IRA project in a county is associated with an increase in Harris’ vote share of 0.028 percentage points in that county and a decrease in Trump’s vote share of 0.026 percentage points when compared with Biden and Trump’s 2020 election performances.

2. Infrastructure investments were associated with a greater impact on vote share than manufacturing investments.

3. The estimated effect on a county with a median number of IIJA, CHIPS, and IRA projects was a better vote share performance of roughly 0.11 percentage points for Harris compared with Biden, and worse performance of 0.10 percentage points for Trump in 2024 compared with 2020.

This analysis builds on previous CAP analysis on the location of private investments and is an early attempt at assessing the electoral impact of the Biden-Harris industrial policy program.² The analysis for this report contains rigorous controls that attempt to account for other factors influencing vote share, such as partisanship, immigration, and local economic and demographic factors. In addition, the authors conducted multiple models and types of analysis to check the robustness of the findings. Still, the analysis suffers from several limitations that restrict the authors’ ability to definitively quantify the electoral impact of the IIJA, CHIPS, and the IRA: The investment datasets do not claim to be exhaustive and include only announced projects rather than projects that have been started or completed. Nevertheless, this research is an important first step at moving beyond exit polls to measure the effects of industrial policy spending and to provide information to build on in the future.

The results of this analysis suggest that voters rewarded the Biden-Harris administration for attempting to improve local economies and create good jobs, but the impact proved modest in the 2024 election. These small gains are likely to raise questions about the relative importance of industrial policy in short-term political strategies—especially because the political impact of these investments also appears modest compared with the electoral returns on other types of federal spending such as additional discretionary funding for state programs in prior elections. (See “Comparing results with other research on spending programs” section below)

The investments made by the Biden-Harris administration spurred a significant increase in private investment in American infrastructure and manufacturing capacity in key industries, created thousands of jobs, and are expected to produce more significant economic gains in the future.³ The electoral effects of these investments could also be more substantial. The political effects of industrial policy investments where projects are underway rather than merely announced could be more significant than those found in this report. The laws were passed in 2021 and 2022; thousands of projects have yet to begin, and even more are still underway, meaning the potential for the IIJA, CHIPS, and the IRA to create quality jobs for the working class and strengthen local economics largely has yet to be realized, assuming disbursement of funds continues uninterrupted.⁴ Future voters may appreciate these impacts.

Former President Joe Biden signed into law three landmark industrial policy packages—the IIJA, CHIPS, and the IRA—to overcome decades of disinvestment in American communities; rebuild the nation’s physical, digital, and utility infrastructure; retake the global lead in advanced semiconductor manufacturing; speed the nation’s transition to electric vehicles (EVs) and clean energy; and create high-quality jobs.⁵ These three policies combine direct public spending with grants, loans, tax incentives, and other financial assistance for private companies to promote key sectors, especially manufacturing, using public investment to “crowd in” private investment.⁶

A previous CAP analysis found that the Biden-Harris administration industrial policy platform successfully achieved its first goal, fostering private investment in communities that have suffered from decades of disinvestment.⁷ Tens of thousands of projects have been announced, and new projects are largely located in communities that lost manufacturing jobs. In fact, 83.2 percent of counties receiving new private investments spurred by the Biden-Harris administration have manufacturing sectors that had shrunk since 2001—a greater share than among all U.S. counties (73 percent). Additionally, these federal investments led to a boom in construction across the country,⁸ with the construction of manufacturing facilities in the United States nearly doubling from 2021 to 2023.⁹ A report from the Brookings Institution found that investments in specific “strategic sectors” are likelier to be in “economically distressed” communities with low employment and income levels.¹⁰

Although the industrial policy program can potentially deliver major benefits for American workers, commentators observed from the beginning that the Biden administration’s bet on industrial policy was electorally risky.¹¹ As a method for winning votes, industrial policy spending suffers from inherent limitations. Even if construction begins soon after the government awards investment recipients, the economic benefits from projects—especially manufacturing projects—may take years to materialize. Although investments in new factories and infrastructure grow local economies, only a small number of local workers will receive the jobs created directly by the project, and voters who do not receive one of these jobs may not associate a better local economy with the investments. Furthermore, the Biden administration’s ability to extract job quality guarantees from recipients was limited as the underlying laws had limited job-quality standards, meaning jobs created may not always be high quality.¹²

Research has long found that U.S. voters care most about national economic conditions and are at best moderately responsive to local economic conditions.¹³ A study by political science researchers Justin de Benedictis-Kessner and Christopher Warshaw analyzed county-level economic conditions from 1969 to 2018 and concluded that a sizable increase in local wage growth of one standard deviation increases the vote share of the sitting president’s party by only one-quarter to one-half of a percentage point.¹⁴ Studies also found that federal spending—independent of local economic conditions—can have some electoral impacts. For example, political scientists Douglas Kriner and Andrew Reeves found that spending increases have to be substantial in order to produce meaningful electoral returns, with an 80 percent increase in federal spending in a county from 1988 to 2008 increasing an incumbent presidential candidate’s vote share by just more than 0.5 percent.¹⁵ These differences may be enough to sway a very close election but require significant investment of political capital from an administration to achieve. Furthermore, in 2024, on the national level, voters struggled with high costs due to the inflationary episode during and after the COVID-19 pandemic, with the result that only 38 percent of Americans felt confident in President Biden’s handling of the economy in May 2024, two months before the former president ended his reelection campaign.¹⁶ Many observers argue that Trump, meanwhile, successfully exploited working-class economic anxiety in his 2024 campaign for president.¹⁷

Workers also must be made aware of the government’s role in supporting these investments. In the 2024 election, the president who signed the investment packages into law was not the candidate on the ballot, further weakening the link in voters’ minds between the investments and their ballot options. The strength of messaging about federal projects in particular plays a crucial role: Researchers Justin Grimmer, Solomon Messing, and Sean Westwood found that in prior elections constituents were more responsive to the total number of messages sent about local federal spending rather than the amount claimed, meaning that voters needed to be frequently informed by the administration about the investments.¹⁸ Indeed, previous CAP research found that even workers on a major subsidized industrial policy project—the BlueOval City EV battery manufacturing plant in Stanton, Tennessee—were largely unaware of the federal support that spurred the project.¹⁹ This suggests that modest political gains should be the expectation.

The authors of this report separately analyzed data on industrial policy investments from two different datasets—the CAP “Biden Administration Investment Tracker”²⁰ and the Biden-Harris White House’s “Investing in America” data formerly hosted at Invest.gov²¹—with elections data from Dave Leip’s Atlas of U.S. Elections²² and local economic and demographic data from the U.S. Census Bureau’s Quarterly Workforce Indicators²³ and American Community Survey.²⁴ By regressing the difference in a county’s vote share for each party’s candidate between 2020 and 2024 on variables measuring the number of projects in that county, along with a selection of controls, the authors estimated how much the share of the vote received by former Vice President Kamala Harris in 2024 increased or decreased compared with the share of votes former President Biden received in 2020—and similarly, how much better or worse President Trump performed in 2024 compared with 2020—associated with each additional IIJA, CHIPS, and IRA project each county received. More details on the methodology and the limitations of the authors’ approach can be found in the Appendix.

According to the CAP analysis, additional industrial policy projects were associated with a higher county-level vote share for Vice President Harris in 2024 compared with President Biden in 2020 and similarly associated with a lower vote share for President Trump in 2024 compared with 2020. Figure 1 shows the marginal change in vote share between 2020 and 2024 for each candidate associated with each additional IIJA, CHIPS, or IRA project, using both CAP and White House investment data. Using CAP investment data and controlling for each county’s presidential election results since 2012, percentage of residents with a college degree, percentage of residents born outside the United States, percentage of Hispanic residents, local unemployment rate, and number of workers in the local labor market, each additional IIJA, CHIPS, or IRA project increased Vice President Harris’ vote share by 0.028 percentage points. Each additional project also decreased President Trump’s vote share by 0.026 percentage points when compared with Biden and Trump’s 2020 election performances. The results are much the same using White House investment data, which found that Harris improved on Biden’s performance by 0.028 percentage points per project while Trump’s declined 0.025 percentage points per project.

Infrastructure investments were associated with a slightly larger effect than manufacturing investments in the CAP data. Figure 2 compares the marginal change in vote share for each candidate from 2020 and 2024 when infrastructure and manufacturing projects are taken into account, using CAP investment data and controlling for local voting, demographic, and economic factors. While each additional infrastructure investment was associated with a 0.031 percentage point better electoral performance for Harris and worse performance for Trump compared with each party’s 2020 candidate, additional manufacturing projects were associated with a smaller effect of only 0.025 percentage points of additional vote share for Harris compared with Biden and 0.014 fewer for Trump in 2024 compared with 2020. This may be the result of several factors: Manufacturing projects may take more time to create new local jobs, and new factories may have less perceived benefit to local voters compared with a new or updated infrastructure project that directly improves roads, internet access, or access to clean water. Results for infrastructure investments using the White House database were statistically significant and similar in magnitude to the CAP dataset, while results for manufacturing investments were not statistically significant. (see Download, Table 3)

The total effect of the IIJA, CHIPS, and the IRA on voting can be estimated for a county receiving a typical number of investments. The median number of projects per county was four; in both datasets, a county with four projects saw an estimated 0.11 percentage point better performance for Harris than Biden associated with the industrial policy program, and a 0.10 percentage point worse performance for Trump in 2024 compared with 2020. (see Figure 3) While this estimate holds for a typical county, some counties may have experienced a greater or lesser difference because they have a different number of projects or other differences in local characteristics.

The results of this analysis indicate that industrial policy investments were associated with a better election performance for Vice President Harris compared with President Biden and a worse performance for President Trump in 2024 compared with 2020; however, the magnitude of the difference is small. How effective were these programs at persuading voters compared with other types of spending?

Researchers have known for decades that increasing federal spending and the number of local projects on which federal money is spent—so-called “pork barrel” spending—can increase electoral performance for incumbent candidates.²⁵ In particular, Democratic candidates are likely to benefit more from spending on projects than from other kinds of monetary transfers.²⁶

A comparison with the existing literature on the electoral effects of federal spending finds that the Biden-Harris industrial policy program was not especially effective at improving electoral performance in the 2024 election. An analysis of data from 1983 to 1990 by economist Steven Levitt and political scientist James Snyder Jr. found that $100 more in federal spending per person in a congressional district on mostly discretionary funding for state programs boosted incumbent vote share by 0.42 percentage points.²⁷ When adjusted for inflation, this equates to 0.17 percentage points for every $100 per person in 2024 dollars, a larger difference than the increase in Democratic vote share of only 0.011 percentage points for every $100 per person in 2024 compared with 2020 using CAP data and a rough comparison of the average county population and value of projects from the IIJA, CHIPS, and the IRA.

An analysis of pork barrel programs in 1984, 1986, and 1988 by political science professor R. Michael Alvarez and economist Jason Saving found that every $100 million of real new spending on these programs was rewarded by an additional 0.15 percentage points more of the vote for incumbent members of the House of Representatives; adjusting for inflation, $100 million in spending in 2024 dollars would equate to 0.046 percentage points more vote share for the incumbent.²⁸ Using the average project cost from the CAP data, IIJA, CHIPS, and IRA spending was associated with a predicted increase in Vice President Harris’ presidential vote share of 1.1 percentage points compared with President Biden for every $100 million spent per county, though with an average of $11.6 million spent per county on new projects, the resulting total advantage for Harris in the 2024 presidential election was far smaller. Infrastructure projects made a more significant difference, at an expected improvement for Harris of 2.1 percentage points compared with Biden for each $100 million in infrastructure projects, though the average spending per county on new infrastructure projects was only $5.6 million. More recently, an analysis by associate professor of economics Emiliano Huet-Vaughn of municipalities located near American Recovery and Reinvestment Act projects in New Jersey found that federal highway investment increased incumbent vote share by 1.5 percentage points in 2012; the state received a total of $570 million in funding.²⁹

Owing to differences in empirical approach, these comparisons are rough at best—in particular, this analysis uses only announced projects, which underestimates the effect each completed project would have on the vote—but give a sense of how the electoral effects of IIJA, CHIPS, and IRA investments compare with some other types of spending.

After the 2024 election, critics were quick to point out that the Biden-Harris administration’s investments in infrastructure and manufacturing, despite their scope, ultimately did not deliver a win for Harris.³⁰ This report offers evidence that the Biden-Harris industrial policy platform likely helped the vice president, but to a relatively small degree. Possible reasons for this small impact include the timing of the projects, the size of the investments, the types of industries they were concentrated in, and the effectiveness of the administration’s messaging.

And of course, industrial policy was but one of many factors that shaped the election. The economic impact of other factors such as the COVID-19 pandemic and the subsequent inflationary episode may have been more important and could have influenced the impact of the IIJA, CHIPS, and the IRA in ways not clear. Furthermore, despite President Biden’s pro-worker rhetoric, gauges of worker well-being such as real incomes for the working class and union density had not reversed their decadeslong stagnation. President Trump specifically used campaign rhetoric reflecting working-class discontent, and some critics theorize that the rift between the Democrats and the working class is due in large part to social issues that new jobs or pro-worker rhetoric were unlikely to overcome.³¹

The Biden-Harris administration made a bet that voters would reward it for industrial policy investments that create good jobs for the working class. By measures of investment volume, the industrial policy platform was successful; the platform will continue to promote local economic benefits for years if it remains uninterrupted. This report finds that, while the investments were rewarded by voters, the effect was small and ultimately did not lead to a Harris victory.

This evidence is important for elected officials and working-class advocates seeking strategies to improve conditions and receive political benefits. It is also useful for researchers looking to investigate the electoral impacts of industrial policy spending. This report offers a starting point that can support future research into how voters react to programs aimed at creating quality jobs and whether other factors affected the effectiveness of the IIJA, CHIPS, and the IRA at influencing voters.

Tables 2–4 can be found in the downloadable spreadsheet available here.

To assess the impact of industrial policy investments on voting, the authors matched data on county-level voting outcomes in the past four presidential elections with the locations of industrial policy projects that had been announced—though not necessarily completed—by government agencies awarding funding and private companies or state governments receiving funding. The difference between each candidate’s vote share for each county in the 2024 election and the share for the candidate of the same party in 2020 was regressed on an investment variable (number of projects in each county in our primary models) and a set of controls to produce the results in this report. Controls include a range of economic, political, and demographic indicators that have been shown to affect local voting results.

Voting outcomes were measured as the difference in performance for a party’s candidate in 2020 and 2024. As a result, the regressions measure the effects of the investment variables and controls on how much better or worse Harris performed compared with Biden, and on how Trump’s performance changed between 2020 and 2024. A separate regression is conducted for each party.

The authors separately analyzed data from two different datasets on industrial policy investments: one compiled by CAP³² and one formerly made available by the Biden-Harris White House at Invest.gov,³³ which other researchers have used for similar analyses.³⁴ Both of these datasets include lists of announced projects receiving funding via IIJA, CHIPS, and IRA with information on the value of the project, its location, and some details on whether the project invests in infrastructure, manufacturing, or another program.

While these datasets are the best sources of information regarding the volume and value of industrial policy investments on a county-by-county basis, they suffer from several limitations that may affect the results of this analysis. Because the datasets were compiled from USAspending.gov, public press releases, industry associations, and news articles, neither dataset claims to be exhaustive, nor does either dataset provide information on whether the projects have been started or completed. Additionally, the White House dataset does not include data from USAspending.gov for projects receiving IRA or CHIPS funding or data from USAspending.gov for IIJA programs valued at less than $100,000. In CAP’s analysis the authors included projects that had been announced as of May 2024 for the CAP dataset and April 2024 for the White House dataset. Using these cutoff dates increases the likelihood that projects were underway and providing economic benefits for communities by the time of the election. However, many of these announced projects were still in the planning stages. The underlying White House database also included projects announced not long before the election; these were excluded from the main analysis to include only projects announced as of April. Preliminary analysis that included these more recently announced projects produced results that were similar but less substantively and statistically significant to those presented in this report, suggesting that actual work on a project may be important for political results. The White House began offering Invest.gov datasets for download beginning in the fall of 2023, but versions prior to the April 2024 version used in this analysis do not include consistent county-level geographic identifiers. Thus, the authors do not perform additional analysis on these releases of project announcements.

Dollar values of public investments in both datasets include just the value of the federal component of funding, regardless of the amount of additional private investment or whether that funding is a grant, loan, loan guarantee, or other funding stream. It is also worth noting that these datasets measure only a specific type of spending—investments made available for infrastructure projects through IIJA and investments in private projects through CHIPS and IRA—although other types of government spending may have had an independent impact. The three investment laws also effectively provided additional funding in the form of tax breaks for private companies, though this is not tracked and therefore not measured in this analysis.

While this analysis was conducted on the county level, economic benefits of projects may spread well beyond county borders, either because of their location within a county or because workers can commute to or take advantage of projects completed in counties outside of their place of residence and voting. The authors also analyzed the data using the sum total of projects within each county and every neighboring county. Results were similar in terms of sign but generally lower in terms of magnitude and statistical significance. Results were weaker still using the large commuting zone groupings of counties originally developed by the U.S. Department of Agriculture.

The CAP and White House datasets include only projects that were mapped by their compilers to specific geographic locations, and their compilers omitted announcements for which no precise location is available. In both datasets, it appears that at least some projects may have been matched to corporate office locations, resulting in a few outlier counties, such as the District of Columbia, with a disproportionately large number of projects. To ameliorate this issue, the researchers omitted the top 5 percent of counties by either value or number of projects. Alternative approaches—such as omitting specific counties, including the District of Columbia and New Castle County (which contains Wilmington), Delaware, that had large numbers of projects due to the presence of corporate headquarters and offices—produced similar results. After removing outliers and projects without geographic locations, the CAP dataset lists 11,753 investments and the White House dataset lists 12,912 investments. Further summary statistics can be viewed in Table 1.

The primary independent variable for this analysis was the number of investments in each county. For the CAP dataset, this is further broken down into total number of infrastructure projects and total number of manufacturing projects. The results of these regressions can be found in Table 2 in the downloadable Microsoft Excel file available above. These regressions are the source for the figures presented in the body of the report. The “Investment coefficient” column shows the difference between 2020 and 2024 vote share for each candidate associated with each additional investment. The results were all statistically significant at at least the 95 percent confidence level except for the effect of the number of manufacturing investments on Republican vote share, which was not significant.

Table 3 and Table 4 in the downloadable file present additional regressions that provide additional analysis and context for the main findings. Table 3 shows that a separate effect for manufacturing and infrastructure investments was not always found: Separate counts for infrastructure investments were statistically significant in the White House dataset and similar in magnitude to results using CAP data while manufacturing investment counts were not statistically significant, unlike in the CAP data.

The analysis was also conducted using the total dollar value of investments in each county. However, this was only statistically significant at a weak confidence level for the White House dataset and not significant at all using the CAP dataset. The discrepancy may be due to the nature of the data on spending, as it was largely derived from USAspending.gov data, which report the value of the federal component of investments but not the sometimes much larger state or private share spent on projects. Further, large dollar value projects may also require more time to plan. The results from conducting this analysis using dollar value of investments per county can be found in Table 4.

Local voting trends, as well as local demographic and economic trends, must be controlled for because all have significant impacts on voting outcomes. According to these controls, Trump performed better in 2024 than in 2020—and Harris worse in 2024 compared with Biden in 2020—in counties that were less college-educated, more foreign-born, more Hispanic, and with less unemployment and a larger local labor market. The authors analyzed voting outcomes on a county-by-county basis, the smallest unit of measurement for which voting results are available. Voting data were sourced from Dave Leip’s Atlas of U.S. Presidential Elections, which includes a county-level dataset of election results. To control for previous voting trends, the authors included the vote share a candidate’s party received in the 2012 and 2016 presidential elections. Demographic controls measured the county’s proportion of residents with at least a four-year college degree; the proportion of residents born outside the United States; and the proportion of Hispanic residents. County-level data on these demographic characteristics were sourced from five-year American Community Survey (ACS) samples.³⁵ Economic controls measure the local unemployment rate and the size of the local labor market in terms of the number of jobs. Unemployment rate was also sourced from ACS data, and the number of local jobs was derived from the Quarterly Workforce Indicators provided by the U.S. Census Bureau.³⁶ Some of these indicators are lagged: Demographic and unemployment data date from 2021 and total local employment dates from 2023, the most recent years available for this data at the time of the analysis. Reporting immigration variables as changes over time rather than levels had no impact on our results. The authors also evaluated a range of other local economic and demographic controls but found that they did not have a statistically significant effect on voting outcomes when combined with the controls used in this analysis, nor did they change the basic results presented in this paper.

Other researchers conducted a preliminary analysis of the effect of investments on voting outcomes based on an early version of the Atlas of U.S. Presidential Elections data, which at the time of their analysis omitted results from a number of counties and Alaska, and the White House investment dataset shortly after the election.³⁷ Their analysis used a different selection of local economic and demographic controls and measured investment using spending per capita, which they found had no effect on voting.