U.S. Library Underserved Population Online Resource Dataset

18 min readJun 5, 2023

Introduction

This project was created to fulfill the final project requirements for the University of Washington’s INFO 572 Introduction to Data Science with Professor Melanie Walsh.

As an MLIS student, I went into this assignment motivated to find comparative data on social services and resources across U.S. libraries, but I could find none. While extensive data exists on details of library staffing, programming, budgets, and more (Public Libraries Survey, 2020), I could find none that audited resource availability for underserved populations (Gray, 2018). Thus, out of my own curiosity and for the value of tracking resource availability nationally, I decided to create a dataset to fill this gap.

Dataset Creation

Over the course of the last four weeks (May 2023), I created my own dataset (U.S. Library Underserved Population Online Resource Dataset) analyzing the presence of online resource pages for underserved populations in the top 100 public libraries and library systems in the U.S. I derived the data on the top libraries ranked by service population size from the 2020 Public Libraries Survey (Library Search & Compare, 2020).

For the other dimensions of the dataset, I personally collected data on the presence of resources for six underserved populations: Non-Citizen and Non-English Speakers, Disabled People, Low-Income and Houseless People, Incarcerated Plus, Veterans, and Seniors.

What is an underserved population?

This term is used in a variety of contexts from library science to public health. The American Library Association describes their list of ten “Special Needs and Underserved Populations” as populations that are “traditionally underrepresented”. The Department of Health and Human Services describes underserved, vulnerable, and special needs populations as “communities that include members of minority populations or individuals who have experienced health disparities” (Serving Vulnerable and Underserved Populations, n.d.). The Health Equity & Policy Lab describes underserved and marginalized populations as “people who experience discrimination of any kind and encounter barriers (e.g., racial, ethnic, gender, sexual orientation, economic, cultural, and/or linguistic) to accessing public health and health care goods and services” (Underserved, Marginalized and Vulnerable Populations, Areas and Facilities, n.d.).

The final list of six populations included in this dataset is not exhaustive and does not encompass all underserved populations in the U.S., but attempts to capture the most commonly cited populations with resource pages on top library websites. Please see the Population Parameters and Definitions section for more information on the selection criteria.

Why does this matter?

This is an especially understudied and underdocumented element of library resources and social services, and this project has the potential to add a valuable new dimension to research in the field.

More broadly, library resource availability for underserved populations is especially important now with threats to social services nationwide (“Hunger Cliff”, 2023; Congressional Republicans’ Legislation, 2023), and the lasting effects of the COVID-19 pandemic on these populations.

Increased vailability of these resources on library websites is just one step that can be taken to make resources accessible and visible for people in these underserved populations.

Research Questions

The following questions were made possible by the creation of this dataset, and, for the most part, answered in the below data analyses.

Which states have the most resources represented on their websites on average? Which states have the least resources represented on their websites on average?
Which regions have the most resources represented on their websites on average? Which regions have the least resources represented on their websites on average?
Which locale has the most resources represented on their websites on average? Which locale has the least resources represented on their websites on average?
Which underserved populations are most represented in online resource pages on library websites? Which populations are least represented?
Which underserved populations had the most outside resources linked instead of internal resource pages? Which populations had the least outside resources linked instead of internal resource pages?
Is there a difference in the number of average resource pages in states with different political affiliations? Is the distribution of population resources different according to political affiliation?

Potential Uses of the U.S. Library Underserved Population Online Resource Dataset

This dataset has the potential to reveal accessibility and resource trends across the nation’s largest libraries. My hope is that this resource can be used by a number of populations, including library workers, researchers, community groups, and underserved populations themselves. My analysis is just the beginning of the analyses that are possible with this dataset, and I aim to expand the dataset in the future to make it even more viable for research and use. (See Future Work)

Internal Reflection

Libraries represented in this dataset have an opportunity here to consider which underserved populations have readily available resources on their websites and which don’t and fix these gaps. There are many lessons to be learned about what resources are currently abundant and which populations are not having resources made available for them, as well as how to best represent these resources.

Examples for Other Libraries

These top 100 libraries, which serve the largest populations in the country, likely serve as examples for smaller libraries and branches that look up to them for indicators of what resource pages should be made accessible, how to organize them, and what to include in population resource pages. Because the dataset can be narrowed down by locale, state, and region, other libraries in similar regions or with similar demographics can utilize this dataset as examples of comparable resource organization and availability for their own libraries.

Resource Expansion

For the general public, this dataset has the potential to highlight gaps on top library websites, and ideally motivate libraries to fill these gaps. One purpose of this project is to accumulate resources for libraries to use as examples for resource development, and ultimately lead to an increase in population resources on library websites across the county. With the filling of these resource gaps comes an increase in resources for these underserved populations. This dataset also serves as a repository of up to 100 examples of each population’s resource page and has the potential to be used by community groups to compare, utilize, and analyze local and regional resources.

U.S. Library Underserved Population Online Resource Dataset Methodology

To access this dataset as a Google spreadsheet, click here.

Collection Methods

Data for this project was collected by myself, Phoebe Brand, a first-year MLIS student at the University of Washington iSchool.

Outside Data

The foundation of this dataset was derived from the 2020 Public Library Survey (Library Search & Compare, 2020). The PLS surveys 9,000 public libraries and 17,000 public library outlets across all 50 states annually, and is carried out by State Data Coordinators from the Institute of Museum and Library Services (Public Libraries Survey, 2020). I first sorted the entire dataset by service area population and then took the top 100 libraries ranked by service population. I also retained two columns from this dataset, Service Area Population and Locale. For the locale column, I cleaned the data so that the three categories of City (City-11, City-12, and City-13) and three categories of Suburban (Suburban-21, Suburban-22, and Suburban-23) were combined into just City and Suburban for the sake of analysis and simplicity.

I manually added columns on location-related data, including State, Region, and Political Affiliation. I used US Census Bureau Regions and Divisions to assign Northeast, Midwest, South, and West to each library (Index of /Geo/Pdfs/Maps-Data/Maps/Reference, n.d.). To assign political affiliation, I assigned Democratic or Republican according to election results from the 2020 election (2020 Presidential Election Results, n.d.).

*Note: Political affiliation is complicated, constantly in flux, and could have been assigned other ways, but for the sake of simplicity I used this method.

Manual Collection of Underserved Population Resource Pages

The remainder of the dataset was collected manually over the course of four weeks. The process of reviewing library websites and determining the presence or absence of resources was iterative and time-consuming, and my approach evolved and improved as I became more familiar with the nature of the websites. A few key changes came about as I began my exploration.

One key development in my process was how to ascertain the presence or absence of resources. I developed the following system in which I could determine this efficiently and consistently:

First, I go to the library’s main page and look for a resource tab (many libraries have this). There, I look to identify resources for underserved populations and mark whether they are present or not.
Next, I explore further menu options and pages for missing populations across the library website.
If I still cannot find resources internally on the library website, I search on Google “Name of Library resource for underserved populations name”.
If this does not yield results, then I feel confident that I can say there are no resources available for that population, and if they are available, they are not readily accessible.
I kept the amount of time I spent per library to under 10 minutes.

In the process of developing this audit method, I also discovered that library websites sometimes had a link to an outside resource (government website, non-profit, etc ), but not a resource page or collection of resources for the population in question. Thus, I added a category of “Outside” resources, indicated on the dataset as “O”, to indicate the presence of an outside link but not a resource page or collection of resources.

Below is a definition for “Yes”, “No” and “Outside” that I used to keep my audit consistent across the wide range of library websites and formats.

“Yes”

Library website page with a list of resources.
More than one resource (not just one link on the website, see “Other”)
Does not have to be the resource variable name(population labels vary widely), and can be a resource list embedded on another population’s resource page if the subsection is labeled.
Example) Oftentimes veteran resources were subgrouped under financial assistance sections, but I count it as a “Yes” for Veterans if it was labeled and more than one resource was included in the subsection.
Example) Oftentimes older adults and senior citizens are included under accessibility pages. I count this as a “Yes” for senior citizens if it was labeled and more than one resource was included in the subsection.

“No”

No resources or resource pages were found on the library website after a thorough audit and search.
Event pages for certain populations were counted as a “No” because they are temporal resources (and oftentimes expired events).
Broken links or faulty links were categorized as “No”.

“Outside”

One link to outside resources
Not a resource page or labeled section for the population. Instead, a singular link to an outside website
Oftentimes these were links to a resource page on another website with thorough regional resources.
Example) Outside resources included city or state websites with various social service resources for low-income, veterans, etc. Oftentimes this was a city-wide 211 site.

Population Parameters and Definitions

One dilemma that I faced while creating this dataset was which under-served populations to include in my survey. Like much of this project, my selection process for narrowing down six populations was iterative.

In the first round of compiling this list, I sought out a specific list of underserved populations in libraries, which I was able to find on the ALA website under “Special Needs and Vulnerable Populations” (Gray, 2018). In order to narrow down to a manageable number of populations for this project and also to those most commonly cited on library websites, I decided not to include Adult New and Non-Readers, Bookmobile Communities, Rural, Native, and Tribal Libraries, LGBTG+, and People of Color. Further time and research are needed to identify resource availability for these groups (see Future Work). I also replaced Spanish from this list with Non-Citizens and Non-English Speakers as those resources tended to be grouped together on the websites I explored, and I wanted to include more than just Spanish speakers. I also decided to add Veterans as a population, because they were oftentimes included on resource menu tabs.

Below is a description of each population included in the final dataset:

Non-Citizens and Non-English Speakers: This includes immigrants, people that are not U.S. Citizens, and those who do not speak English fluently. Resources for this population were oftentimes labeled for “Immigrants”, “Non-English Speakers”, “New Readers”, “En Español” and more, and included resources like ELL classes, guides studying for citizenship tests, and resources for asylum seekers, to name a few.
Disabled People: The ADA’s legal definition is “a person who has a physical or mental impairment that substantially limits one or more major life activity” (What Is the Definition of Disability under the ADA? | ADA National Network, n.d.). Resources for disabled people were oftentimes listed under pages like Accessibility Resources, with resources like large-print books, talking books, and homebound services.
Low-Income and Houseless People: This population includes people that are living in poverty, are low-income, and/or are houseless. Low-income and houseless people may face food insecurity, housing insecurity, and/or depend on social services for health care and basic necessities, and resource pages oftentimes present resources to address these issues.
Incarcerated Plus: The Incarcerated Plus population includes currently incarcerated, formerly incarcerated, and families of incarcerated people. I decided to include the last one as I discovered several websites that included this. Resources for Incarcerated Plus ranged in labeling, including “Decarcerated”, “Transitional” and “Reentry”.
Veterans: The U.S. Department of Veterans Affairs describes a veteran as a person who “served in the active military, naval, or air service, and who was discharged or released therefrom under conditions other than dishonorable” (Utilization, n.d.).
Senior Citizens and Older Adults: This population includes those who are senior citizens (over the age of 65) or older adults who may benefit from services like accessibility assistance, technical help, and community.

See Bonus Section for examples of the different ways these populations were represented and labeled on library websites, as well as further examples of resources.

Challenges and Limitations in Collection and Analysis

I want to recognize that this is not a perfect auditing system, as the presence or absence of a resource page is a judgment call by myself as a data collector and open to error.

For that reason, I am including a submission page here for readers to submit edits to the presence or absence of library website resources. If I missed a resource for a library, please submit it here and I will add it ASAP.

What is this dataset not measuring?

This dataset is not meant to be a value assessment of library websites or library resources. There are several reasons a library may not have resources for certain populations listed, including but not limited to budgeting, staffing constraints, community priorities, political pressures, and judgment calls on whether or not online resources are the best presentation of resources for a certain population.

It is also not meant to be a ranking of the best library websites or the worst, as there are infinite possibilities of how libraries design their websites and make their priorities. Rather, this dataset is meant to be an opportunity for comparisons, collaboration, and growth in resource availability for U.S. libraries.

Lastly, this project only measures the presence or absence of resources, not the quality, breadth, or effectiveness of the resources. More detailed and focused qualitative work would be needed to accomplish that research.

Labeling of Resource Pages

Population resources are not always named explicitly for the populations they are meant to serve and are called different things across different websites. For example, for formerly incarcerated people, the NYPL calls their resource page “Connected”, and does not explicitly state that it is for formerly incarcerated people in the title. This is something I quickly became aware of while auditing websites, and learned how to look for commonly used labels and names. However, this does increase the possibility of overlooking a population resource page. See the Bonus Section for the array of labels and resource-organizing systems I found while making this dataset.

Limited Dataset Size

Due to time constraints, I was only able to audit and analyze the top 100 library websites ranked by service population. Thus, in some categories like regions and states, there are limited numbers of libraries per category. As seen in the map below, the grey states did not have libraries represented in the Top 100 library systems by service population. Regionally, this means that the West North Central, East North Central, and East South Central regions have particularly less representation in this dataset, which may skew their wider regional data as well.

Map of the Number of Libraries Analyzed Per State

In contrast to the under-represented regions, some states were highly represented in the top 100 libraries. Below is a list of the top 5 states:

California: 18 libraries

Florida: 13 libraries

Texas: 8 libraries

Maryland: 6 libraries

Washington: 6 libraries

In contrast, 12 of the 31 states represented in this dataset only had one library represented.

Computational Method

For analysis, I used R Studio with several packages for visualization and mapping purposes:

dplyr
ggplot2
usmap
plotly
tidyr
stringr
tidyverse

All visualizations below use a color package with colors that are legible and distinguishable for readers who are colorblind. (See cbp1 package)

Findings and Insights

Average Resources by State

The top 5 states according to the average number of internal resources pages (“Yes”) are:

Massachusetts: 6.0

Colorado: 5.33

New York: 5.25

Kentucky: 5

Washington DC: 4.0

Hawaii, Missouri, and Utah were tied for the bottom three, with 1 average resource each.

Some interesting and more statistically significant trends are revealed when we organize these states by region.

Regional Analysis

The Northeast has the highest average number of internal resource pages, at 4.11. The Midwest has the lowest average, at 2.25. This is a wide gap in average resource availability, and further investigation into the reasoning behind this trend could be very interesting. Some possible explanations for this may be government support of libraries and their funding, political affiliations, regional demographic needs, and pressure from patrons to make resources available.

Locale Analysis

There is not a large difference in the average number of internal resource pages between City and Suburban libraries. Suburban libraries averaged 2.64 internal resource pages, and City libraries averaged 3.21. This gap may be due to the more diverse demographics of urban areas, funding, and population need.

*Note: There was only one rural library in the dataset, which I removed for relevancy.

Underserved Population Resource Prevalence

The above chart shows the number of libraries out of the 100 sampled that include an internal resource page (“Yes”) for each underserved population. Non-Citizen and Non-English population internal resource pages were the most prevalent, at 82 libraries. In close second is the Disabled population, at 79. Incarcerated Plus had the lowest rates of representation on internal resource pages with only 24 libraries including resource pages for this group.

Non-Citizen and Non-English is a combined population category, and probably had the widest range of types of resource pages that counted towards a “Yes”.

In contrast, very few libraries (24) contained websites for the Incarcerated Plus population. As in the above analyses, this may be due to library priorities, the demographic needs of their service populations, or political pressures.

The above chart shows the number of libraries out of the 100 sampled that include an internal resource page (“Yes”) or a linked outside resource page (“Outside”). I decided to only include and analyze the “Outside” category for this aggregated population resource page section in order to see which populations had the highest rates of “Outside” resources being included for them. The population with the highest number of “Outside” resource pages was Low-Income and Houseless people, with 13 “Outside” resource pages, significantly increasing total resource page availability from 35 to 48. In contrast, only one website included an “Outside” resource for Incarcerated Plus.

Resource Availability by Population

Below is the breakdown of resource pages by “Yes” (inclusion of an internal resource page), “No” (no resource pages), and “Outside” (link to an outside resource page) for each underserved population. This reiterates the above analysis with a new visualization to examine the population by population data and trends.

Political Affiliation

As discussed in the Collection Methods section, I used 2020 Presidential election data to assign “Democratic” or “Republican” to each state. Below is the assignment of each state, with blue indicating Democratic and red indicating Republican.

Below, you can see the average number of internal resource pages for Democratic states and Republican states. Democratic states had a higher average number of internal resource pages, at 3.39, while Republican states had 2.3.

The following chart includes only Democratic states, so as to see the breakdown of internal population resource page prevalence. In all population categories, Democratic States had higher rates of internal resource pages than both the aggregate of all libraries and the Republican states. In some cases, there were much higher rates than Republican states, such as in the case of Senior Citizens, with 50% of Democratic states vs. 28% of Republican states including internal resource pages.

The Republican libraries follow a similar pattern in resource availability as Democratic states and aggregated dataset but with much lower rates of resource availability in each category. You can hover over each population’s bar to see the contrasting rates of resource page availability.

One explanation for these gaps by political affiliation may include pressures from governments on library staff. One needs only look at the rates of book banning in southern and Republican states to see the effect that political agendas can have on library policies.

Future Work

My hope is that this dataset can be expanded in the future to include more libraries and more resource categories. Some categories I would like to add are library budgets, the presence of accessibility buttons on the website, the presence of social workers in libraries, and LGBTQ+ resource pages.

There is also the potential for much more data analysis on this existing dataset (June 2023). For example, it would be interesting to cross-compare political affiliation with budgets (available through the PLS) to see if Republican states are less well-funded and if that may explain resource availability.

Bonus Section: Examples of Library Resources

The following section contains just some of the stand-out examples and unique resources I found while creating this dataset.

Senior Citizens

Several library websites included homebound services like book-by-mail services for senior populations and other homebound populations.

Disabled People

The Harris County Public Library provides adaptive equipment including “Comfort Kits” for people for whom “the library is just a little too much to handle”.

Low-Income and Houseless

I noticed that resource pages in Southern states tended to include religious ministries.
While I did not count this as a resource page for the affected population, several libraries had reading lists about “Big Topics” like homelessness.
Some libraries, including the Boston Public Library, had specific pages for Teens or Youth experiencing Housing Insecurity.
Many library websites had “Career” pages which I did not count as low-income or houseless pages unless there were explicit sections designated for those groups.
The San Francisco Public Library has a Job and Careers Section entitled “Resources for Groups”, which includes specified job resources for groups like “Reentry”, “50+ Job Seekers”, and “LGBTQIA”.

Incarcerated Plus

Incarcerated Plus services went under a wide range of names, including “Decarcerated”, “Transitional” and “Reentry Services”.
Like houselessness, some libraries had reading lists for children about incarceration.

Veterans

When searching for Veteran resource pages, I oftentimes stumbled upon Veteran History pages instead.

Miscellaneous:

Kern County Library has a cumulative resource page called “Overcoming Challenges”, which includes sub-sections for Seniors, Disabled People, and Health Services.
LA County Library has a “Cultural” resource menu tab with resource centers for populations like a Black Resource Center and Chicano Resource Center. They also have a drop-down menu for Local History by neighborhood.
The St. Louis County Library includes a resource page with links to community resource maps by the Clark-Fox Family Foundation, with interactive resource maps by category.

References

2020 presidential election results. (n.d.). Retrieved June 4, 2023, from https://www.cnn.com/election/2020/results/president

Adaptive Resources. (n.d.). Retrieved June 4, 2023, from https://hcpl.net/adaptive-resources

Alboukadel. (2018, November 18). GGPlot Colors Best Tricks You Will Love. Datanovia. https://www.datanovia.com/en/blog/ggplot-colors-best-tricks-you-will-love/

Big Topics, Little People: Homelessness — Gwinnett County Public Library. (n.d.). Retrieved June 4, 2023, from https://www.gwinnettpl.org/kids/big-topics-little-people-homelessness/

Community Resources | St. Louis County Library. (n.d.). Retrieved June 4, 2023, from https://www.slcl.org/content/community-resources

Congressional Republicans’ Legislation: 22% Cuts That Would Harm American Families, Seniors and Veterans | OMB. (2023, April 20). The White House. https://www.whitehouse.gov/omb/briefing-room/2023/04/20/congressional-republicans-legislation-22-cuts-that-would-harm-american-families-seniors-and-veterans/

Decarceration Resources | Riverside County Library System. (n.d.). Retrieved June 4, 2023, from https://rivlib.info/services/decarceration-resources

Geeza, M. (n.d.). LibGuides Home: Libraries and Accessibility: Accessible Websites. Retrieved May 7, 2023, from https://libguides.ctstatelibrary.org/dld/accessibility/websites

Gray, J. (2018, October 16). Special Needs & Underserved Populations [Text]. Advocacy, Legislation & Issues. https://www.ala.org/advocacy/advocacy-university/special-needs-underserved-populations

Home. (n.d.). 2–1–1 Metro Chicago. Retrieved June 4, 2023, from https://211metrochicago.org/

“Hunger cliff” looms as 32 states are set to cut food-stamp benefits. (2023, February 27). https://www.cbsnews.com/news/food-stamps-snap-benefits-cuts-march-2023/

Index of /geo/pdfs/maps-data/maps/reference. (n.d.). Retrieved June 4, 2023, from https://www2.census.gov/geo/pdfs/maps-data/maps/reference/

LA County Library. (2023, May 31). LA County Library. https://lacountylibrary.org/

Library, A. L. A. (n.d.). LibGuides: Library Service to Persons with Disabilities: Web Accessibility Guidelines. Retrieved May 7, 2023, from https://libguides.ala.org/libservice-disability/web-accessibility

Library Search & Compare. (2020). http://www.imls.gov/search-compare

Overcoming Challenges — Kern County Library. (n.d.). Retrieved June 4, 2023, from https://kerncountylibrary.org/overcoming-challenges/

Public Libraries Survey. (2020). http://www.imls.gov/research-evaluation/data-collection/public-libraries-survey

Reentry Services. (2020, September 15). https://www.bklynlibrary.org/outreach/justice-initiatives/reentry-services

Resources for Groups | San Francisco Public Library. (n.d.). Retrieved June 4, 2023, from https://sfpl.org/locations/main-library/jobs-careers-center/jobs-and-careers-resources/resources-groups

Resources for Youth Experiencing Housing Insecurity. (n.d.). Retrieved June 4, 2023, from https://www.bpl.org/resources-for-youth-experiencing-housing-insecurity

Richter, S., Bell, J., Jackson, M. K., Lee, L. D., Dashora, P., & Surette, S. (2019). Public Library Users: Perspectives of Socially Vulnerable Populations. Journal of Library Administration, 59(4), 431–441. https://doi.org/10.1080/01930826.2019.1593711

Serving Vulnerable and Underserved Populations. (n.d.).

Transitional Services Division | www.buffalolib.org. (n.d.). Retrieved June 4, 2023, from https://www.buffalolib.org/services/transitional-services-division

Underserved, Marginalized and Vulnerable Populations, Areas and Facilities. (n.d.). Health Equity & Policy Lab. Retrieved June 4, 2023, from https://www.healthequityandpolicylab.com/underserved-populations-areas-and-facilities

Utilization, O. of S. A. D. B. (n.d.). VA.gov | Veterans Affairs [Homepage]. Retrieved June 4, 2023, from https://www.va.gov/OSDBU/

What is the definition of disability under the ADA? | ADA National Network. (n.d.). Retrieved June 4, 2023, from https://adata.org/faq/what-definition-disability-under-ada