A public library may be worth more to a neighbourhood when people can reach its doors easily. Research on American funding referendums suggests that physical access, rather than the size of a tax bill alone, helps explain whether residents vote to support libraries.
In an analysis of voting precincts in Washington state, having a library building within the precinct was associated with approximately 13% greater likelihood of support for library funding. Support declined by roughly 4% for every additional mile of travel distance in a more fully adjusted model.
The study, published in PLOS One on 1 October 2026, draws on actual referendum outcomes rather than simply asking people how much they like libraries. Its central finding is that proximity to a shared public facility can be reflected in financial decisions at the ballot box.
Researchers Margo Gustina and Robert P. Berrens, both affiliated with the University of New Mexico, analysed library votes recorded over more than a decade. They combined a national dataset with detailed Washington state records that allowed them to examine distances between voting areas and library buildings.
The findings do not mean that building a library automatically increases electoral support by a particular percentage. The research is observational, and communities with nearby libraries can differ from more distant places in ways that are difficult to measure.
Nevertheless, the pattern adds a practical dimension to debates about public infrastructure. The question is not only which services a library offers, but also who can conveniently reach them and whether that access shapes willingness to finance the institution.
What the researchers measured
In many American communities, residents periodically vote on property-tax measures that fund public libraries. These ballots provide a record of people deciding whether to support a public service when a financial contribution is involved.
The authors assembled a national library referendum dataset covering voting events from 2010 to 2021 for the main national models. Their wider compilation contained 1,679 voting events involving 802 libraries.
Not every event had sufficient information for every calculation. The compilation included 1,136 events with enough information to estimate tax incidence, relating to 479 unique libraries, while individual regression specifications used different eligible subsets.
For the principal comparison, the authors report 982 library system-year observations nationally. This unit represents a library system in a particular voting year, not 982 individual voters.
A second dataset followed Washington state precinct-level outcomes from 2010 to 2022. The authors report 5,380 precinct-year observations in their analytical dataset, offering a finer view of where votes were cast.
The researchers linked these voting records to information about estimated household tax burdens, income, education, political voting patterns, settlement characteristics and exposure to extreme heat.
Washington also supplied the crucial spatial detail: whether a precinct contained a library and how far the precinct was from the nearest library building. The national dataset was less precise about that relationship.
Statistical models estimated how the proportion of votes supporting library funding changed alongside those characteristics. The models adjusted for several contextual factors, but they cannot remove every possible source of confounding.
A nearby building made a substantial difference
The most striking result concerns a simple geographical fact: whether a library is located within the same voting precinct. The authors' overall summary reports an increase of approximately 13% in the likelihood of a Yes vote when a precinct contains a library.
One of the Washington models estimated an approximately 12% increase for the same characteristic. The exact size varies by model specification, so the result is best understood as a consistent association rather than a fixed effect.
Distance also mattered. In the adjusted Washington analysis, every additional mile between the voting precinct and a library was associated with approximately 4% lower likelihood of support.
Across statistically significant estimates, the distance relationship ranged from about 3% to 8% lower support per mile. The researchers describe the overall adjusted relationship as roughly 4% per mile.
The authors also found that about 6% of Washington precincts were closer to a library outside the system whose funding was being decided. In one model, this mismatch was associated with a 16% lower likelihood of support.
That finding is consistent with a potential disconnect between the service people can reach and the system they are asked to finance. It does not establish that voters were deliberately protesting against a particular funding arrangement.
Importantly, the study does not show that every library should be placed within walking distance of every resident. It shows that the geography of access is a material part of the observed funding relationship.
Taxes mattered, but they were not the whole story
Funding referendums often involve a specific property-tax levy. A basic economic expectation is that a larger household cost should make people less willing to vote in favour.
The researchers estimated the annual tax incidence of each proposal, meaning the likely household contribution implied by the levy and local property values. This allowed them to compare financial cost with other features of the community.
In the national models that included the relevant controls, each additional US dollar in estimated tax burden was associated with roughly 0.1% lower likelihood of a Yes vote. The direction was broadly consistent with economic expectations.
That figure should not be read as a universal price response for individual voters. It is an estimated relationship in aggregated referendum data, and the paper reports variation between model specifications.
The Washington models also generally found a negative relationship between the estimated tax burden and support. Yet spatial variables, especially the presence of a library in the precinct, showed comparatively large associations.
Household income complicated the picture. Wealth and tax incidence are closely related because the levies are linked to property values, making it difficult to separate their statistical effects.
In one Washington model, an additional US$10,000 in household income was associated with an approximately 3% higher likelihood of voting Yes. The authors caution that overlapping financial variables can complicate interpretation.
The broader lesson is that a funding proposal's price is only one part of its public value. Location, access and the existing pattern of service provision can help explain why similar tax requests receive different responses.
A public facility can serve more than one purpose
The study also explored whether extreme weather conditions were associated with support for library funding. This reflects the possibility that libraries provide shelter, cooling and other services beyond books and lending.
In the national models, each additional day with a heat index above 100 degrees Fahrenheit was associated with an estimated 1.7% higher likelihood of a Yes vote in one reported specification.
The heat relationship was not uniform across the analyses. The authors report different directions for the association in the national and Washington datasets, making it unsuitable as a simple prediction.
One possible explanation is that communities exposed to severe heat see greater value in publicly accessible indoor spaces. Another is that heat exposure is correlated with other geographic or economic characteristics.
The study cannot determine whether voters consciously considered a library as a cooling centre. It measures relationships between local conditions and voting, not the reasons stated by individual residents.
Still, the finding encourages a wider view of community infrastructure. A library can provide internet access, study space, information, social contact and, in some places, relief during difficult weather.
Putting an estimated value on proximity
The researchers also attempted to translate voting behaviour into an annual willingness-to-pay estimate. This asks what financial value can be inferred from how communities respond to proposed library taxes.
In selected Washington library systems, estimated annual household willingness to pay ranged from approximately US$25 to US$65 in the central point estimates. Broader estimates ranged from about US$4 to US$99.
These amounts are model-derived valuations, not money that residents actually handed over in a separate experiment. They depend on assumptions about tax costs, income and the relationship between voting and preferences.
The authors also estimated how library value declines with distance. In one analysis using statistically significant results and many precincts, an additional mile was associated with about US$33 less annual household value.
That figure is not a national benchmark and should not be converted into a price tag for an individual library. It applies to the particular analytical setting and modelling assumptions used in the study.
Voters must interpret what that rate means for their own finances. The researchers therefore estimated tax incidence using available property and levy information instead of observing every person's actual cost.
Even with those cautions, the monetary exercise reinforces the central spatial result. Being farther from a library was associated with a meaningful reduction in the value inferred from local funding decisions.
Why the South African comparison requires care
The study is based on American voting systems in which library funding can be placed directly on local ballots. South African public libraries operate within different intergovernmental and municipal funding arrangements.
It would therefore be misleading to apply the study's 13% proximity figure or its dollar valuations directly to South African library budgets. The voting institutions, travel patterns and financing structures are not the same.
Yet the question of physical access is highly relevant wherever public facilities are unevenly distributed. Travel costs, transport availability and opening hours can determine whether a nominally free service is practically usable.
A library that is several kilometres away may impose costs in time and transport, particularly for people without private vehicles. A centrally located facility can be easier to use repeatedly, even when the services offered are identical.
For planners, the research suggests that distance should be examined alongside footfall, population size and operating expenditure. A single large branch may be efficient in some respects but less accessible to people at the edge of its catchment.
That is a planning consideration, not a finding directly tested in South Africa. Local decisions would need evidence about actual journeys, service demand, community preferences and the costs of different branch networks.
These potential benefits require their own evidence. The referendum results show an association between access and political support, not a measured improvement in educational outcomes or employment.
Important limitations of the evidence
The strongest limitation is that the analysis is observational. The researchers did not randomly allocate library buildings to precincts, change their locations or randomly vary tax rates across comparable communities.
Areas with nearby libraries may have different settlement patterns, civic traditions, income distributions or political histories. Statistical adjustments help, but they cannot guarantee that every relevant difference has been captured.
The data also describe aggregate voting outcomes. They do not link each ballot to a person's library use, income, age, reasons for voting or awareness of the nearest branch.
This creates a risk of ecological inference: a relationship measured for precincts should not automatically be treated as a relationship for every individual voter within those precincts.
Another limitation is selection into voting. People who participate in a local funding referendum may differ systematically from residents who do not vote, particularly in income, property ownership or engagement with public services.
Not all library funding votes were equally observable. The national dataset was assembled from multiple reporting sources, and the availability of tax information restricted which events could enter some models.
The two datasets also differ in geographic precision and composition. Washington's detailed distance measures support stronger spatial analysis, but findings from one state need not transfer unchanged to every American region.
Some estimated relationships shifted across specifications, including the role of extreme heat and the strength of the distance association. The study reports these differences rather than a single uniform result for every setting.
Finally, willingness-to-pay estimates depend on model choices and assumptions. They should be read as analytical approximations, not as verified market prices for access to a public library.
What the findings mean for public decisions
The clearest implication is that a public building's location can shape how residents value the institution that occupies it. The visibility and convenience of a library may matter alongside its collections and programmes.
Where officials face pressure to consolidate branches, proximity should therefore be considered as a potential source of public value, rather than treated only as a property or transport issue.
Equally, the findings do not justify building new facilities without examining demand, cost and alternatives. Mobile libraries, transport connections, opening hours and digital access might change how people experience distance.
The research provides a reason to test those options, not a definitive ranking of them. A local evaluation could compare travel times, service usage, costs and resident experiences before deciding how to distribute resources.
Funding debates often present public services as a choice between affordability and benefit. The referendum evidence shows why that framing can be too narrow: a facility's value may depend on whether residents can realistically use it.
The authors' results place access at the centre of that relationship. A nearby building was associated with stronger support, while increasing distance generally corresponded with weaker support.
Whether the same pattern would appear in other countries or under different funding arrangements remains an open empirical question. Future research could examine changes following branch openings, closures or transport improvements.
For now, the evidence suggests that the geography of a public service is part of its social and political value. What communities can reach may influence what they are prepared to sustain.
Source Information
Study title: Voters reveal public library value: Evidence from US referendums
Authors: Margo Gustina and Robert P. Berrens
Journal: PLOS One, volume 21, issue 10, article e0351352
Publication date: 1 October 2026
Study design: Observational analysis of actual US public library funding referendums using regression models
Study period: National data mainly 2010 to 2021; Washington state data 2010 to 2022
Sample and units: 982 national library system-year observations and 5,380 Washington precinct-year observations; these are aggregated voting records, not individual survey participants
Original study DOI: 10.1371/journal.pone.0351352
Funding and disclosure: The authors report partial funding from the Institute of Museum and Library Services. Gustina disclosed that she is a librarian with a favourable disposition toward public libraries.
