How To Actually Map The Conservative Institutional Network
Most people think mapping the conservative movement means making a chart of think tanks and PACs. It is not that simple. The real work is figuring out where the money moves, who sits on whose board, and which policy language gets reused across half a dozen organizations over a ten year period. When I started doing this kind of mapping a few years back, I assumed I could just download some datasets and build a graph. That did not work. The connections are deliberately opaque. Foundations layer through multiple vehicles. Think tanks spin off affiliates that quietly coordinate funding. A single donor might route money through a state level group, a national super PAC, and an educational nonprofit all at the same time, and none of those filings explicitly connect them. What you are really looking for is not a list of organizations but a dependency map. Every institution in this space depends on a small number of funding pipes and a handful of key intermediaries. The Heritage Foundation, the Federalist Society, the Club for Growth, FreedomWorks, Americans for Prosperity, the Bradley Foundation, the Scaife network, the Koch network, the DeVos network, the Mercer family connections, the OxyGen Group, the Stanford Center on Philanthropy and Civic Society, and dozens of state level counterparts form the skeleton. But the skeleton is held together by relationships that never appear in IRS Form 990 filings. Here is the practical approach I use now, after burning through about eight months trying it the naive way.
Start with foundation grant data. CandidGuide, the Foundation Center database, and IRS Published Activity filings will give you the raw grant flow. Export everything and normalize the recipient names yourself because the same organization appears as "Heritage Foundation Inc," "Heritage Foundation," and sometimes just "Heritage Fdn" depending on who wrote the check. I spent two weeks doing fuzzy string matching in Python with the rapidfuzz library before I accepted that no automated tool would get it clean enough on its own. From there you need board overlap analysis. This is where most people stop and miss the actual wiring. If three people sit on the boards of Heritage, Goldwater Institute, and the Institute for Justice simultaneously, those are not three independent organizations. They are a coordinated node. Pull board listings from 990-PF schedules and cross reference them manually. I built a simple SQL database with tables for person_id, org_id, and role_type. The query to find triply linked individuals took me maybe twenty minutes once the data was loaded. The insight it gave me was significant. Then trace the policy text. This is the part that feels like overkill until you have actually done it. Take key legislative proposals and white papers from the Federalist Society, Heritage, Cato, and the State Policy Network. Run them through a simple cosine similarity check using TF-IDF vectorization. You will find that the same policy language migrates from one organization to another with only minor edits. The State Policy Network especially operates as a conduit. It is not just a coalition of state think tanks. It is a content distribution system. What gets written at the national level gets packaged and pushed down to affiliates in Wisconsin, Florida, Ohio, and Texas. I saw this repeatedly when I was tracking education voucher legislation in 2022. A bill introduced in Florida had nearly identical provisions to one filed in Ohio three weeks later, and both traced back to a SPN template document.
There are tools that help here. The OpenSecrets database gives you donor flows for PACs and Super PACs. The League of Conservation Voters has good corporate giving data. But the real gaps are in the dark money space where 501(c)(4)s and 501(c)(6)s do not have to disclose their donors. I found a workaround for this by filing FOIA requests with the IRS for the EIN databases of specific organizations and then cross referencing the disclosed officer lists with publicly available professional bios and LinkedIn profiles. It was tedious and it required writing a small script to handle the PDF extractions, but it got me names that no grant database would show me. One particular edge case stood out. I was trying to trace funding for a state level advocacy group in Georgia that claimed independent expenditures in a congressional race. Their IRS filing showed no donors above a certain threshold. I ended up finding the actual funding source by following a board member who also sat on the advisory council of a midwestern foundation, which had previously given a grant to a national organization, which had received a multi million dollar grant from a family foundation. That chain was invisible in any single database. It only appeared when you forced the connections together. The mapping itself works best when you visualize it with Gephi or similar network analysis software. Use force directed layouts and color code by funding type rather than by organization type because that reveals the real patterns. You will quickly see that the conservative network has a star topology around a small number of family foundations and a hub and spoke structure around certain legal and advocacy organizations. The Federalist Society, for example, functions as a talent pipeline and ideological coordinator rather than just a legal organization. Its network position is different from Heritage or Cato even though they appear similar on the surface. There are serious limitations to this approach. You cannot map what does not leave a paper trail. The more an organization relies on in-kind services, volunteer labor, or private donations routed through intermediaries, the less your model will reflect reality. I have seen cases where the primary funding mechanism was simply a wealthy donor hosting events and asking attendees to contribute directly to causes, with zero formal organizational structure to trace. Your map will show empty space where the actual activity happened.
Get the Full Details
Another problem is that conservative institutions evolve faster than public data reflects. A think tank might shift its strategic focus over eighteen months, absorbing staff and projects from another organization, rebranding its website, and reissuing its materials under a slightly different name. The IRS filings lag behind these changes by quarters. I learned this the hard way when I published a preliminary map that listed several organizations as separate entities when they had effectively merged their operations but not their legal structures. The map was technically correct but practically misleading. If you need something quicker and less granular, the State Policy Network publishes its own member directory and annual reports which give you a decent starting framework. The Heritage Foundation also maintains extensive public documentation of its partnerships. These sources are useful for orientation but they will not show you the hidden connections. For that you have to do the manual work I described. The final step most people skip is validation. After you build your network map, pick five random nodes and manually verify each connection using at least three independent sources. If you cannot verify the link, mark it as uncertain in your dataset. Your map will be better for having honest uncertainty markers than for pretending every connection is confirmed. I keep a separate spreadsheet with columns for connection_type, source_evidence, confidence_level, and last_verified_date. It makes the whole thing reproducible and defensible when someone challenges your findings, which they will.
This is not a complete guide to understanding American conservative institutional power. It is a description of how to build a working map of it. The data will always be incomplete. The relationships will always be partially hidden. But a reasonably accurate picture is possible if you treat it as an engineering problem rather than a research project and accept that you will spend far more time on data cleaning and verification than on the actual mapping.
