NOTE-042: Bonica, "The Mothership Vortex" (Democratic spam PACs)

Contribution

This is an investigative newsletter post. It follows the money through FEC filings from a cluster of Democratic-aligned PACs to one digital fundraising vendor. It puts a single number on the “efficiency” of spam-style small-dollar fundraising, and argues that the Democratic Party should cut ties with such vendors and set efficiency standards.

Key insight

If a PAC network’s disbursements are traced, very little of what it raises from individuals is passed on as contributions or transfers to candidates and party committees. The rest is consumed by the vendor, texting services, payroll and more fundraising. On this reading the “necessary evil” defence (“it works”) mistakes gross receipts for political output.

Assumptions

These are premises and authorities.

FEC data are the sole evidence: committee summaries (individual contributions), Schedule B disbursements, and Schedule A receipts, pulled through the OpenFEC API.
The “network” is a hand-coded list of 35 committee-ID entries (31 distinct IDs) in the script. “Clients” is a second list of 17. How membership was decided is not stated beyond “Mothership-linked”. End Citizens United’s co-founding by Berlin and Starnes is the one specific link the text states (unsourced).
“Reaching campaigns” is operationalised in code as Schedule B rows in purpose category CONTRIBUTIONS or TRANSFERS to committees outside the two lists. Independent expenditures (Schedule E) are not queried by the script.
Mothership’s founding story (2014; founded by the former DCCC digital director Greg Berlin and his deputy Charles Starnes; “churn and burn”) is stated without sources.

Key results

What the post reports, attributed to the author:

$678M raised from individuals by the core network since 2018 (excluding candidate clients).
$159 M p ai d b y t h e n e tw or k t o M o t h er s hi p, o u t o f$ 282M Mothership received from all clients.
$22.5 M t o M ess a g eD i g i t a l LL C . A b o u t$ 150M “consulting/fundraising” and $70 M p a y r o l l .$ 19M from Progressive Turnout Project to Shawmut Services for canvassing.
“At most” $11M to candidates, campaigns or national party committees, giving 11/678 ≈ 1.6%; “approximately half” of it to the DNC, DCCC and DSCC.
Mothership’s clients also include House Majority PAC and Jaime Harrison. “None of this, as far as I can tell, is illegal.”
The infographic adds per-client payments to Mothership: PTP $37.8 M, S t o pR e p u b l i c an s$ 27.2M, NDTC $22.9 M, C H C B O L D$ 17.0M, Harrison $16.6 M, E C U$ 16.2M, Ossoff $11.9 M, an d o t h er s . I t a l sos h o w s in t er - P A C t r an s f er s in t o P T P (f r o m S t o pR e p u b l i c an s$ 25.9M, Progressive Takeover $13.1 M, D E M T u r n o u t 2024$ 8.4M) and the PTP and Stop Republicans spending panels described above.

Claims

id	claim	strength	support
C1	The network raised ≈$678M from individuals since 2018	the author’s own data analysis; not reproducible from the repo	computed from committee_summary CSVs that are not published (script lines 278–283, 526)
C2	The network paid Mothership $159 M, o u t o f$ 282M from all clients	the author’s own data; partially reproducible in principle	Schedule B search on recipient “MOTHERSHIP”; the `cycle` parameter the script passes is not a documented Schedule B parameter, so the multi-cycle coverage is unverified; the query spans 2016–2026 while the text says “since 2018”
C3	At most $11M reached candidates and party committees	the author’s own data; the bound’s direction is unsupported	code sums out-of-network CONTRIBUTIONS/TRANSFERS from a per-committee sample capped at 1,000 rows, largest first, ≥$1,000, and (as documented) latest period only; IEs are excluded
C4	Efficiency is 1.6%, and “98 cents” of each dollar goes to consultants and operations	arithmetic on C1 and C3 (11/678 = 1.62%); the “98 cents” gloss overstates	the author’s own chart puts $23.2M of PTP spending in field operations
C5	$22.5 M M ess a g eD i g i t a l;$ 19M Shawmut; $150 M co n s u l t in g / f u n d r ai s in g;$ 70M payroll	the author’s assertion from data; no code in the repo for the first two; the keyword regex for the $150M (e.g. “text”, “email”, “digital”, “donation”) may overlap with the others	script lines 536–548
C6	About half of the $11M went to the DNC, DCCC and DSCC	assertion; contradicted by the infographic ($4.0M)	none in the repo
C7	Mothership’s principals co-founded End Citizens United in 2015	assertion	no source given in the post
C8	Nothing here is illegal	the author’s hedged judgement	”as far as I can tell”

Method

The pipeline, as committed in mothership_complete_api_analysis.R:

Search Schedule B for recipients matching “MOTHERSHIP” (lines 206–244), passing cycle=2016,…,2026. The current OpenFEC swagger lists no cycle parameter for /schedules/schedule_b/, and says results “default to the most recent two-year period” when two_year_transaction_period is absent. Whether the API ignored or honoured cycle in August 2025 could not be tested.
Join payments to locally stored committee-summary CSVs for 2018–2026 by committee ID (lines 278–290). These files are not in the repo. Line 318 then references top_100_clients, which is never defined.
Fetch committee details, incoming transfers (Schedule A, ≥$10,000) and outgoing disbursements (Schedule B, recipient_committee_id_null=FALSE, which is not a documented parameter, sorted -disbursement_amount, min_amount=1000, set_limit=10 pages of 100). No transaction period is passed, so the documented default is the latest period only. write_csv is called at line 416 before readr is loaded (line 440), so that tryCatch block reports failure even though its data frame has already been assigned.
Classify disbursements by regex on the description (lines 536–548). Sum CONTRIBUTIONS and TRANSFERS to committees outside the two hand lists to get “Total spent on outside candidates and committees” (lines 566–577).

Concepts

fundraising efficiency rate: the author’s measure, (contributions + transfers to outside candidates and committees) ÷ (individual contributions raised). It excludes independent expenditures, field and canvassing spending, and advocacy, by construction.
“churn and burn”: high-volume, urgency-driven small-dollar solicitation that trades long-term donor relationships for short-term revenue.
core network / connected PAC: the hand-coded list of committees the author treats as Mothership-linked, as distinct from ordinary clients.

Connections

It follows the author’s earlier post documenting the spam tactics themselves (linked, not read). Don Moynihan (c06) links a different Bonica post on shutdowns. The subject overlaps a long-running journalistic literature on “scam PACs”. The post cites none of it, and I have not checked it.

Bearing on the record

It carries nothing for ML practice and no instruction for the Anthology of the SOTA. As a reading-record item it is a worked example of a headline ratio whose numerator and denominator come from differently scoped queries. It is the shape ANTH-DP-010 describes: the 1.6% is the sentence that travels, and the published repo does not carry it.

Limitations

Not reproducible as published. Inputs are missing, the script has a fatal undefined variable, and several reported figures have no code.
A mismatch in scope between numerator and denominator is likely. The denominator is summary data for every cycle from 2018. As documented, the numerator’s disbursement pull covers only the latest period and is capped. The infographic’s Stop Republicans panel ( $17.8 M t o t a l s p e n d in g, a g ain s t$ 27.2M paid to Mothership) is consistent with the numerator seeing only part of the data. This is an inference from the code, the documentation and the chart; it was not run.
The definition of “reaching campaigns” excludes independent expenditures and field operations. A PAC that spends on IEs or canvassing for candidates scores zero for that spending.
The infographic and the text disagree ( $282 M v s$ 159M attributed to the network; $4.0 M v s \approx$ 5.5M to the party; network membership).
No response from the named firms is included. The network lists are hand-coded without stated criteria.

Open questions

What is the efficiency ratio when every cycle is queried with full pagination, and with IEs and field spending reported as separate categories? That run would settle C1–C4.
How does this network compare with Republican and other Democratic small-dollar operations measured the same way? Without a comparison, “1.6%” has no reference point.

Corrections to the seeded skim

The dossier says replication is the “only deeper work” and “the claims are fully captured”. Auditing the repo (see Method and Limitations) finds that the headline numbers are not reproducible from it as published. The script reads committee_summary_2018–2026.csv files that are not in the repo. It references an undefined object (top_100_clients, line 318), which halts a non-interactive run whenever those files are present. It contains no computation at all for the $22.5 M (M ess a g eD i g i t a l),$ 19M (Shawmut), $150 M /$ 70M split or “about half to the DNC, DCCC and DSCC” figures.
The dossier gives the $11 M a s an u pp er b o u n d (" p a sse d a t m os t ab o u t$ 11M”), following the post. The code gives no upper bound. Outgoing disbursements are fetched sorted by amount descending, capped at 10 pages × 100 records per committee, with min_amount=1000 and, as documented, only the most recent two-year period, because no two_year_transaction_period is passed. Candidate contributions, which are small, are the records such a cap drops first. The $11M is therefore, if anything, a truncated count.
The dossier omits the infographic, which contradicts the text in four places. (a) “What the Network Gets: $282 M in r e v e n u e p ai d t o M o t h er s hi p ", w h er e a s t h e t e x t s a y s t h e n e tw or k p ai d$ 159M and $282 M i s a l l c l i e n t s . (b) " W ha tt h e P a r t y G e t s :$ 4.0M in direct donations”, whereas the text has “approximately half” of $11 M g o in g t o t h e t h r ee na t i o na l co mmi tt ees . (c) T h e ba r c ha r t co l o u r sH o u se M aj or i t y P A C, A M E R I P A C, C H C B O L D P A C an d L e t A m er i c aV o t e P A C r e d a s " cor e n e tw or k ", w h er e a s t h eco d e l i s t s t h e f i r s tt h r eeo n l y a sor d ina r y c l i e n t s an dd oes n o tl i s tt h e f o u r t ha t a l l . (d) I t s S t o pR e p u b l i c an s p an e l s h o w s t o t a l s p e n d in g o f$ 17.8M ( $10.8 M t r an s f er s +$ 7.0M fundraising and media), while the same graphic shows Stop Republicans paying Mothership $27.2 M an d t r an s f er r in g$ 25.9M to Progressive Turnout Project. The spending panels evidently cover only part of the data.
The dossier’s check-list asks whether independent expenditures count. The infographic’s PTP panel does count “Direct Candidate Support (Contributions & IE)” ( $4.0 M, 6$ 23.2M (36%). The text’s “98 cents goes to consultants and operational costs” therefore absorbs field operations that the author’s own chart separates out.

Status	Read
Version	2
Paper	LIT-060 — The Mothership Vortex: An Investigation Into the Firm at the Heart of the Democratic Spam Machine

nucleation

Explorer