This page is the public evidence annex for Israel, Lobbying and Western Policy: What the Evidence Shows in the US and UK.
Its purpose is not to repeat the main article.
It makes the test visible.
The question we actually tested
The investigation did not ask whether foreign governments, lobbying organisations, donors, campaign groups or companies can ever influence politics.
They can.
The question was narrower:
For a specific policy episode, what evidence connects a specific intervention to a specific governmental effect?
That requires more than showing that actors knew each other, met each other, shared preferences or spent money.
The causal chain
The frozen evidence model separates six stages:
| Stage | Meaning | What it does not prove |
|---|---|---|
| Connection | An identifiable relationship exists | Influence |
| Access | The actor can reach a policymaker or institution | That the access changed anything |
| Intervention | The actor sought an identifiable outcome | Success |
| Transmission | The request, argument or information reached the relevant decision process | That the decision changed because of it |
| Behavioural/process effect | An observable decision-maker or institutional response followed | That this effect was necessary or sufficient |
| Policy contribution | Evidence supports a bounded contribution to the outcome | Control over the wider policy system |
This separation is the central methodological safeguard in the investigation.
What counts as stronger causal evidence?
The strongest positive findings generally contained at least one of the following:
- an official decision-maker explicitly attributing the decision to the external request or contribution;
- a formally bilateral policy instrument that the two governments demonstrably co-produced;
- documentary evidence showing that submitted material entered the decision process and altered its content;
- a clear before-and-after behavioural change that followed the intervention and survived prior-position and alternative-explanation tests.
The weaker cases often contained real advocacy and access but lacked the later attribution bridge.
Temporal precedence
For every serious causal claim, the intervention had to occur before the relevant decision effect.
A statement made after a decision could show an organisation’s position.
It could not cause an earlier decision.
We therefore distinguished:
- intervention before policy formation;
- intervention after policy formation but before the final decision;
- contemporaneous intervention;
- and post-decision reaction.
Prior-position control
A politician agreeing with an organisation after receiving its support does not show that the support created the view.
The investigation therefore searched for earlier:
- manifesto commitments;
- votes;
- speeches;
- government policy;
- party positions;
- comparable decisions.
Where the target already held the relevant position, later alignment was treated as weaker causal evidence.
Alternative pathways
Every case was tested against competing explanations, including:
- prior party policy;
- law and regulation;
- government legal advice;
- national-security interests;
- bilateral diplomacy;
- humanitarian conditions;
- economic interests;
- public opinion;
- parliamentary pressure;
- government turnover;
- allied policy;
- and external events.
A lobbying theory did not win simply because lobbying was present.
Negative controls
The investigation deliberately retained cases in which:
- an actor asked for something and failed;
- the government delayed or narrowed the requested policy;
- an actor had access but policy moved in the opposite direction;
- policy changed after a change of government;
- or an alternative institutional pathway explained the outcome.
Without those cases, successful alignment can easily be mistaken for causal power.
The US evidence tranches
The US investigation separated four actor/mechanism classes rather than treating them as one network.
| Tranche | What is established | What is not established |
|---|---|---|
| AIPAC / United Democracy Project | Electoral targeting, large independent expenditures, candidate support/opposition, policy advocacy | Purchase or production of policy outcomes |
| Christians United for Israel | Mass mobilisation, lobbying, political access and repeated policy advocacy | Purchase, control or independent production of investigated US policy outcomes |
| Defence / security industry | Assistance-to-procurement-to-commercial-benefit pathway; lobbying and industrial reinforcement | Contractor-specific causation of investigated foreign-policy outcomes |
| Israeli-state bridges | Extensive institutional access and specific bounded causal contributions in some episodes | General Israeli control over US policy |
The distinction between the Israeli state and non-state advocacy organisations is mandatory.
Evidence of Israeli-government contribution cannot automatically be credited to AIPAC, CUFI or another organisation.
A positive US example: Iron Dome emergency funding
The 2014 emergency Iron Dome appropriation provides a relatively strong state-to-state example.
The White House publicly connected the additional $225 million appropriation to Israel’s request for additional Iron Dome funding.
That does not show Israel was the only cause.
Congress and the administration had independent alliance, security and domestic political reasons to act.
But it does provide documentary evidence that the Israeli request contributed to the particular decision.
The result is therefore a bounded contribution, not a finding of general control.
UK core-case table
The UK case universe was frozen before the final causal synthesis.
| ID | Policy episode | Principal competing pathway | Frozen non-state advocacy result |
|---|---|---|---|
| UK-C01 | 2030 UK-Israel Roadmap | Direct bilateral state co-production | NOT DEMONSTRATED for non-state advocacy |
| UK-C02 | Anti-boycott Bill | 2019 Conservative manifesto baseline; parliamentary process | Intervention evidenced; initiation causation constrained by prior policy |
| UK-C03 | Jerusalem embassy review | Truss’s prior position; later government reversal | NOT DEMONSTRATED |
| UK-C04 | UNRWA funding pause/resumption | Allegations triggering pause; independent review and reforms preceding resumption | LFI contribution NOT DEMONSTRATED |
| UK-C05 | Gaza ceasefire / UNSC voting | Resolution language, humanitarian conditions, hostages and allied diplomacy | Actor-specific contribution NOT DEMONSTRATED |
| UK-C06 | ICC jurisdiction | Change of government and legal/policy stance | Advocacy causation NOT DEMONSTRATED |
| UK-C07 | Arms-export suspension | Formal export-control/IHL review | Advocacy contribution NOT DEMONSTRATED |
| UK-C08 | West Bank sanctions | Settler violence/incitement, sanctions criteria and allied action | LFI/Yachad contribution NOT DEMONSTRATED |
| UK-C09 | Upgraded FTA suspension | Israeli-government conduct and Gaza policy, explicitly cited by government | Non-state contribution NOT DEMONSTRATED |
| UK-C10 | Recognition of Palestine | Pre-existing Labour policy plus July conditionality and diplomatic strategy | Organisation-specific contribution NOT DEMONSTRATED |
| UK-C11 | Settlement-trade measures | Legal/occupation and settlement-policy pathway; implementation still developing at cutoff | Advocacy pathway plausible; decision-level contribution NOT DEMONSTRATED |
UK aggregate boundary
Across these 11 selected core cases:
0/11 establish a complete independently evidenced non-state advocacy organisation → attributable behavioural change → UK policy outcome chain.
This count is descriptive of the selected corpus.
It is not an estimate of the probability that UK lobbying ever matters.
Why UK-C02 is a useful control
The anti-boycott episode shows why a chronology can overturn a simple influence story.
CFI and LFI both intervened in the parliamentary process.
But the underlying objective had already appeared in the 2019 Conservative manifesto.
That means later lobbying could potentially affect drafting, emphasis or parliamentary handling, but it cannot be treated as the origin of a policy that already existed.
The Bill then failed to become law.
Both the prior commitment and the failed outcome must remain in the causal record.
Why UK-C07 is different
The September 2024 arms-export suspension contains a much more explicit institutional mechanism.
The government publicly said the decision followed a review and an assessment that there was a clear risk certain military exports might be used in serious violations of international humanitarian law.
That gives the analysis a traceable legal and administrative pathway.
Advocacy for restrictions still matters as political context.
But an actor-specific causal claim requires evidence showing that the advocacy altered the review, timing, scope or final decision.
The frozen record does not establish that bridge.
The matched comparator design
Four comparator networks were selected:
- Saudi Arabia
- United Arab Emirates
- Qatar
- Taiwan
They were not chosen because they are politically or morally equivalent to Israel.
They were chosen because they provide foreign-policy relationships in which similar mechanisms can be observed:
- high-level access;
- security cooperation;
- arms and defence relationships;
- economic investment;
- diplomatic requests;
- political advocacy;
- successful outcomes;
- and governmental resistance.
The comparison therefore tests mechanisms, not actor identity.
Comparator case table
| Case | Frozen result |
|---|---|
| SA-US-01 | NOT_DEMONSTRATED |
| SA-US-02 | NOT_DEMONSTRATED |
| SA-US-03 | NOT_DEMONSTRATED |
| SA-UK-02 | NOT_DEMONSTRATED |
| UAE-US-01 | NOT_DEMONSTRATED |
| UAE-US-02 | ESTABLISHED — bounded contribution |
| UAE-UK-01 | ESTABLISHED — bounded contribution |
| QAT-US-01 | ESTABLISHED — bounded contribution |
| QAT-UK-01 | ESTABLISHED — bounded contribution |
| TW-US-01 | NOT_DEMONSTRATED |
| TW-US-02 | NOT_DEMONSTRATED |
| TW-UK-01 | ESTABLISHED — bounded contribution |
Totals:
- 5/12 ESTABLISHED
- 7/12 NOT_DEMONSTRATED
- 0/12 lobbying pivotality established
- 0/12 general foreign-policy control established
What made the positive comparator cases stronger?
The strongest comparator cases contained formal attribution or co-production.
The UAE-UK Sovereign Investment Partnership was created and expanded through an explicit framework overseen by the UK Office for Investment and Mubadala.
Qatar’s US Major Non-NATO Ally designation was publicly connected by US officials to Qatar’s contributions to US security objectives.
The UK-Qatar defence relationship contains a formally joint Typhoon squadron and government-to-government training arrangements.
These are not assumptions based on access.
They are documented institutional mechanisms.
Comparator negative controls
The comparator corpus also includes:
- congressional resistance;
- veto conflict;
- legal constraint;
- policy delay;
- partial outcomes;
- alternative arrangements;
- and requests that did not produce the preferred route.
The 2019 Yemen War Powers episode is one example.
Congress passed a resolution intended to end unauthorised US military involvement in the Saudi-led Yemen conflict. President Trump vetoed it. The Senate failed to override that veto.
The episode cannot be reduced to either “Saudi Arabia controlled US policy” or “Saudi Arabia had no influence”.
It demonstrates competing institutions and policy forces producing a contested outcome.
Cross-network result
After matching the target cases against the comparators, several broad mechanisms turn out not to be unique:
| Mechanism | Israel-linked corpus | Comparator corpus | Result |
|---|---|---|---|
| High-level foreign-state access | Present | Present | Not distinctive by itself |
| Non-state advocacy / representation | Present | Present in varying forms | Presence does not establish causation |
| Defence and security relationships | Present | Present | Not distinctive |
| Economic interests | Present | Present | Incentive does not equal attribution |
| Policy alignment | Present | Present | Alignment does not prove production |
| Formal bilateral co-production | Present | Present | Stronger causal evidence when directly documented |
| Government resistance / failure | Present | Present | Contradicts automatic-compliance models |
| General policy control | Not established | Not established | Unsupported in both corpora |
Structurally distinctive mechanisms
Two mechanisms remain structurally distinctive in this corpus:
- AIPAC/UDP domestic US electoral intervention
- CUFI-style domestic mass mobilisation
Their broader causal significance remains unresolved. That distinctiveness must not be converted either into proof of exceptional control or into proof of equivalence with comparator networks.
Evidence labels
The final graph uses descriptive evidence labels rather than political scores.
DIRECT_DOCUMENTARY
An official record directly documents the relevant mechanism or attribution.
DIRECT_BEHAVIOURAL
Observable behaviour changes in a way that can be tied to the intervention.
STRONG_CIRCUMSTANTIAL
Multiple pieces of evidence support the mechanism but the attribution is not direct.
LIMITED_CIRCUMSTANTIAL
Evidence is compatible with the mechanism but significant alternatives remain.
ACTOR_SELF_REPORT
The actor claims credit or describes its own intervention.
This can prove what the actor says it did.
It does not independently prove causal effect.
SECONDARY_ONLY
The proposition depends on secondary reporting because primary material was unavailable.
NOT_EVIDENCED
The required proposition is absent from the frozen corpus.
CONTRADICTED
Evidence conflicts with the proposed causal claim.
What “NOT DEMONSTRATED” means
This phrase is easy to misuse.
It does not mean:
the influence definitely did not happen.
It also does not mean:
the influence probably happened but was hidden.
It means:
the evidence required to establish the causal contribution was not present in the frozen corpus.
That distinction is essential to the investigation.
Aggregation cannot repair a missing bridge
A repeated pattern of meetings is a repeated pattern of meetings.
A repeated pattern of donations is a repeated pattern of donations.
A repeated pattern of policy alignment is a repeated pattern of alignment.
Eleven weak attribution chains do not become one strong attribution chain when added together.
The cross-case synthesis therefore preserved every case-level NOT_DEMONSTRATED finding rather than upgrading them through repetition.
Network-level findings
The combined graph supports these propositions:
- Israel-related political access and advocacy in the US and UK are substantial and documented.
- AIPAC/UDP electoral intervention is substantial and documented.
- CUFI mobilisation and lobbying are documented.
- Defence-industry commercial and lobbying incentives are documented.
- Specific Israeli-state causal contributions can be established in bounded episodes.
- The selected UK cases contain repeated non-state advocacy but no complete independently evidenced non-state policy-production chain.
- Comparable foreign networks also show access, state bargaining, success, resistance and failure.
- The strongest causal findings across networks occur where formal attribution or co-production exists.
- Automatic compliance and general foreign-policy control are not established.
- Absence of a demonstrated causal bridge is not proof of zero influence.
Network-level non-findings
The evidence does not establish:
- systematic purchase of US or UK policy;
- a financial quid pro quo;
- automatic compliance with Israeli preferences;
- a single unified Israel-linked command structure;
- deterministic campaign-spending effects on policy;
- universal coordination among the investigated organisations;
- general Israeli control over US or UK foreign policy;
- or complete independence of US or UK policy from Israeli or advocacy influence.
Those are deliberately different propositions.
Small-sample boundary
The case corpus was designed for causal testing, not prevalence estimation.
The investigation therefore cannot answer questions such as:
- “What percentage of US foreign policy is influenced by Israel?”
- “How often does lobbying change UK policy?”
- “Which foreign country is most influential?”
Those questions would require a different research design.
The evidence here can show what happened in selected episodes and what mechanisms the evidence supports.
How to challenge this investigation
The useful way to challenge a finding is to identify a missing piece in the causal chain.
For a case currently marked NOT DEMONSTRATED, evidence could change the result if it showed, for example:
- a decision-maker explicitly saying the intervention changed the decision;
- internal records showing an advocacy request altered policy drafting;
- a demonstrable position change after exposure that cannot be explained by the prior baseline;
- or a process record showing that the intervention was pivotal to timing, content or implementation.
If that evidence appears, the case should be reopened.
The purpose of freezing an evidence graph is not to make its conclusions permanent.
It is to make later changes visible.
Publication boundary
This annex and the main article report the investigation as frozen at 18 September 2026.
Future evidence should be added through a new version rather than silently inserted into the v0.1 analysis.
That protects the distinction between:
what the evidence showed at publication
and
what later evidence may show.