Libraries in Community Systems is a study of public and tribal library value within their local contexts of overlapping and intersecting socio-ecological systems. This component, "Library Value from Public Funding Referendums", examines value and preferences as revealed by voters in ballot initiatives on library taxes. The data come from a variety of sources described in detail in the Data Source documents in the National and Washington case study component files (see below) and in each component's wiki (with source links!). Conceptually, the datasets are synthesizing library service, voter behavior, geographic and spatial (Washington only) characteristics, partisanship, demographic and economic characteristics in voting community, and climate-related hazards. There are a few data challenges common to both datasets - some are common to all studies using Public Library Survey data. I'll begin with those specific to working with referendum data: - No central collection of vote events - Some states collect public funding vote data as part of their state version of the Public Library Survey (or "annual report"), but not by rate - Comparing vote events requires calculating a "tax incidence" per household, but taxation structure is different depending on state and local laws, and rates are not published consistently, even on ballots - Tax incidence, whether property assessment or consumption sales based, are all mixed up with income: income (in part) determines taxes, and it also determines the value of money to you which creates real problems in estimating money value of the library - Total votes cast is necessary for weighting when comparing across locations - Precinct level vote data reveals more rich and nuanced information, but very few states make that information easy to extract - Precinct identifiers change over time requiring harmonization across data sources and years - Spatial mismatches between precinct and Census boundaries produce imperfect estimates of Census gathered community characteristics Most of these challenges can be overcome with dogged determinism and large time expenditures. The calculation of tax incidence, though, can't be eased if the rate information isn't available. Further, if the tax mechanism is a sales tax, it is significantly more difficult to attempt a household estimate based on income and household make-up assumptions of consumption, which can then be compared to property tax levy states which are assessed at the household level. For this strain of research query to be robust, state coordination would need to include collecting tax levy rates in property tax based referendums, which could then be merged with vote event precinct level vote counts. The library data, as collected in the annual Public Libraries Survey, presents granularity challenges. In Georgia, all public library systems report annual service provision and use statistics at the library branch (outlet or building) level. These aren't reported federally, it is one of the only states that collect these details, and Georgia libraries don't put operational library funding on public ballots. Washington state has a clear and useful way of accessing ballot language and per precinct vote counts across all the years of our study. Unfortunately, library systems only report hours, square footage, and weeks open at the branch level consistently during those same years. So, where we have granular library performance statistics, we have no voter preference data, and where we have granular voter preference data, we have only system (often county) level performance statistics. Even so, analysis of these data revealed many insights. Details of these can be read in the pre-print "Voters Reveal Public Library Value" [link coming soon]. Further, these datasets are entirely novel. Our hope is that their construction has been documented thoroughly enough that other researchers will use them in their own research and build on this strain of valuation that has only just begun.