En geografisk analyse af afstanden til offentlige biblioteker i København. Projektet undersøger tilgængeligheden af bibliotekstjenester på tværs af byens forskellige kvarterer.

Distance as an Access Barrier for Library Usage

Kasper Spindler | University of Copenhagen | April 2026

1 Introduction

Library usage does not just depend on whether libraries exist, but on how accessible they are. Distance can act as a practical barrier to physical libraries, and municipalities where people on average live further away from libraries should expect a lower amount of physical book borrowings.

The raw results follow this expectation: municipalities with longer average distance to a library or access point tend to have lower physical loans per capita. The pattern becomes weaker in some specifications, which is not surprising because loans are measured at municipality level while access is built from postcode populations, libraries serve across municipal borders, and borrowing also varies with urban structure, local library resources, digital alternatives, and demographic composition.

Further analysis should be done using microdata that compares loan behaviour between users and non-users of libraries, while correcting for demographic differences.

2 Data

The analysis combines three groups of data. First, library locations come from a cleaned list of Danish public libraries and access points, where addresses are geocoded using DAWA. Second, postcode population and postcode-municipality links come from Statistics Denmark tables POST1NR and POST1A.

Third, library outcomes and library-system indicators come from Statistics Denmark’s national library register: FSBIB1, SBS1, BIB6, BIB8 and IBIB1A. These tables cover public-library activity and resources, including physical loans, digital use and municipal library indicators.

Municipality-level demographic variables come from De Kommunale Nøgletal and include population density, age structure, education and unemployment. The final dataset is constructed at the municipality level and contains 98 Danish municipalities.

3 Methodology

To approximate the average distance each citizen in a municipality has to the nearest library, I construct a population-weighted average distance from postcode city/residential centres to the nearest library or library access point.

First, each library or access point is geocoded using DAWA. Then, a representative location of the population in each postcode is calculated as the median coordinate of all addresses within the postcode. This is used instead of the geometric centroid of the postcode area, as postal areas are irregular, and their centres are not always representative of where people actually live. For each postcode, I then calculate the distance to the nearest geocoded library or access point.

Some postcodes have land areas in more than one municipality. To simplify this, postcode populations are allocated to municipalities using the POST1A postcode-by-municipality data together with the 2024 postcode population data. The municipal average distance is then calculated by weighting each postcode’s distance to the nearest library/access point by the share of the municipality’s population living in that postcode.

I also calculate a library catchment population. Here, each postcode’s population is attached to the nearest library/access point, and these postcode populations are then summed for the libraries/access points located in each municipality. This gives an alternative population denominator for loans per capita, as some people may live in one municipality but use a library in another.

The main model is kept as a simple descriptive relationship between physical library use and average distance:

\[ \log(\text{PhysicalLoansPerCapita}_m) = \alpha + \beta \log(\bar{d}_m) + \varepsilon_m \]

The dependent variable is physical loans per capita from “Folkebibliotekerne”. The main coefficient of interest is $\beta$, which measures whether municipalities with longer average distance to libraries/access points also have lower physical library use. Municipality-level demographic controls are not used as the main model, because they do not necessarily describe the population behind the postcode-level access and catchment measures.

4 Results

The main result is the simple relationship between average distance to the nearest library/access point and physical loans per capita. I keep this as the main model, because municipality-level demographic controls are not a clean match for the population behind the access measure when postcodes and catchments cross municipal borders.

In the main model, a 1% increase in average distance is associated with about 0.086% fewer physical loans per capita. The relationship is statistically significant in the main specification, but weaker across the alternative specifications.

Interpreted literally, living 5 km from the nearest library/access point instead of 1 km is associated with about 13% fewer physical loans per capita in the main log-log model.

Table 1 — Main model
Specification (log-log) Coefficient Std. error p-value
Physical loans per capita, distance only −0.086 0.037 0.020
Table 2 — Alternative specifications and robustness checks
Specification Coefficient Std. error p-value
Catchment denominator, average distance −0.027 0.049 0.577
Catchment denominator, serving distance −0.095 0.052 0.069
eReolen comparison outcome −0.047 0.040 0.240
Excluding Region Hovedstaden −0.045 0.027 0.088
Share of population more than 5 km away −0.027 0.107 0.805
The alternative models test whether the relationship changes when using a catchment denominator, a serving-distance measure, a digital-loan comparison outcome, a non-capital-region sample, or a distance-band measure. These are treated as robustness checks, not as the main model.

5 Figures

Figure 1
Map of library access and postcode distance to nearest library
Dots mark postcode centres; open circles mark geocoded libraries/access points.
Figure 2
Population distance bands by municipality
Municipalities sorted by the share living within 0–2 km.
Figure 3
Physical loans per capita vs. average distance to nearest library
Each point is a municipality, sized by population; the fitted line shows the relationship between average distance and physical loans per capita.
Figure 4
Regression coefficient on log average distance — robustness across specifications
Coefficients with 95 % confidence intervals.