Screen the evidenced drivers of renegotiation pressure on a signed PPP contract: demand against forecast, bid aggressiveness, delivery and scope pressure, incomplete contract areas, regulatory change, external shocks, the election cycle, private-party financial stress, authority capacity and disputes in progress. One status per driver, a band set by the worst driver, and the levers that reduce the pressure inside the existing contract.
v1.0 — September 2026
Ten drivers taken from the renegotiation evidence (Guasch's World Bank study of Latin American concessions, the follow-up ITF paper and the PPP Reference Guide) and from the product's completeness rules, each screened from entries the team can evidence. Statuses are In order, Watch or Pressure; a blank entry reads Unanswered and holds the band, and an answer that says a fact is not known is read at the least favourable answer it could turn out to be. The band is the worst driver, never an average. Paste the contract text and the 45 completeness tests propose which areas the contract leaves open, for you to confirm.
The profile axes are the product's: revenue model, delivery structure and financing decide which of the 45 completeness tests apply and whether the demand driver applies at all.
Paste the agreement, schedules included. The 45 completeness tests of the product run over the text and propose, for each applicable area, "Present" where the mechanism was found or "Mark for review" where it was not. Nothing is filled in until you accept it. A mechanism not found is not proof it is absent; the contract may name it differently. The text stays in this browser tab.
Enter what the team can evidence. Each driver's rule is stated in the Methodology. A blank entry reads Unanswered and no band is suggested until it is made. Where a fact is not known, say so (the "Not known" answer, or the Not known box under a figure): the driver is then read at the least favourable answer the fact could turn out to be. Figures are read strictly: digits, with a point for decimals; a comma, text, a unit, a negative or a figure out of range is refused and named. A % sign is accepted in a percentage field, and a leading zero or a trailing point is read as written (05 is 5).
The 45 mechanisms a PPP of this profile usually has, in the eight areas of the product's completeness rules, each with why it matters. Mark each Present, Absent or incomplete, or Don't know. Rows the profile excludes are greyed out and not counted. An incomplete area is where renegotiation starts: the parties fill the gap by bargaining instead of by the contract.
| Driver | Status | Basis |
|---|
For each driver at Watch or Pressure: what the authority can do inside the existing contract, before anyone proposes reopening it.
The band above is a suggestion computed from the entries. A named person records the decision, and no decision is exported without a name; an override does not change the suggestion, it sits beside it in the export. The block clears whenever the screen is re-run or an entry is changed.
The Renegotiation Risk Screen asks a contracting authority's team to evidence the conditions under which PPP contracts get reopened, and returns one status per driver, a band, and the levers available inside the existing contract. It is screening-level. It does not predict that a renegotiation will happen, does not say whether one would be justified, and does not value anything. Renegotiation is sometimes the right answer to a genuinely incomplete contract; the screen is about seeing the pressure early enough to act inside the contract first.
Three sources. The empirical work on renegotiation, above all Guasch's World Bank study (2004) of about a thousand Latin American concessions awarded between 1985 and 2000, which found that thirty per cent were renegotiated (fifty-five per cent in transport and seventy-four per cent in water, against under ten per cent in electricity), that the average renegotiation came 2.2 years after award, that sixty-one per cent were initiated by the operator and twenty-six per cent by the government, and that the incidence was far higher where the award had been on the lowest tariff (sixty per cent, against twenty-nine per cent for awards on the highest fee), where no regulatory body existed (sixty-two per cent against twenty-three), and under price-cap regulation and regulation by means rather than by objectives. His regressions also carry an election-year variable, and the follow-up ITF paper of 2014 reads the recent Latin American cases, dominated by additional works and construction-phase changes, against the political cycle. The World Bank PPP Reference Guide's guidance on managing PPP contracts, which treats renegotiation as a contract-management risk to be managed through the contract's own mechanisms. And the product this series draws on, PPP Contract Review & Operating Map v1.0.0, whose 45 completeness rules define what a contract of a given profile usually contains; each one the contract lacks is an area the parties will later fill by bargaining, which is this tool's reading of Guasch's finding that contracts regulating by means and detailed obligations were renegotiated more than those regulating by objectives.
The demand driver uses the same forecast-accuracy framing as the Infrastructure Pre-Feasibility Toolkit's Demand Assessment: the empirical over-estimate ranges from Flyvbjerg, Holm and Buhl for rail and from Bain and Standard & Poor's for toll roads, with the HM Treasury Green Book's principle of optimism-bias adjustment. Here they serve as the base rate against which an actual-to-forecast ratio is read. The Green Book guidance covers capital cost, duration and operating cost rather than demand; only its principle of adjusting for optimism bias is borrowed.
Every threshold above is the tool's own screening convention, chosen to separate the ordinary from the worrying, and every one is stated so that it can be argued with. None is a finding from the literature; the literature supplies the drivers, not the cut-offs.
The screen never prints anything more comfortable than the entries support. Every answer to every entry falls into one of four readings, and the table below is generated from the one the code runs. A real answer is read as recorded. An answer that says the fact is not known is read at its least favourable: every real answer the fact could turn out to be is tried, and the worst driver status is kept; the basis says so, and no fact is printed for it. A blank entry reads Unanswered: the driver prints no status better than that (unless every possible answer already gives Pressure), and no band is suggested until it is made. A figure is read strictly, as digits with a point for decimals inside its range; a comma (which can be a decimal comma or a thousands separator), text, a unit, a negative or an out-of-range figure is refused (a % sign in a percentage field, a leading zero and a trailing point are accepted, so 05 reads 5), marked on the field and named, and treated as blank.
An answer that says a fact is not known is read at the least favourable real answer it could turn out to be, so it never gives a better band than that answer would; it never gives a knockout, because no knockout was recorded.
No score, no weighting, no average. The count of drivers at each status is shown for orientation only; the band is the worst driver, the tool's convention, because a single driver (a demand shortfall, an election, a lender default) can be enough to bring a request. The band is a suggestion that a named person accepts or overrides with a reason; both appear in the export. No override is accepted while no band has been suggested, nor on a band set by a recorded knockout: a knockout gives High whatever else shows, so only acceptance is recorded, and the confirmation line names the knockout and counts the entries still outstanding. A High band set only by entries that are not known, read at their least favourable, says so in its headline and lists those entries.
The 45 tests are the product's completeness rules, carried over unchanged: for each mechanism, the area, why it matters, what to consider and when it does not apply, together with the pattern the product runs over operative text to see whether the mechanism is there. Each test carries the product's applicability (every profile; structures that build; project-financed; availability-based) and its severity (high or medium). The product's own description of this layer applies here too: it carries a material false-positive rate by design, nothing in it is critical, and every finding quotes the reason it might legitimately not apply. That is why the paste-text step proposes "Mark for review", never "Absent", for a mechanism it did not find.
The incomplete-areas driver reads the checklist as follows: a Don't know is read as Absent or incomplete, its least favourable answer; then any applicable high-severity mechanism absent gives Pressure, whatever else is unanswered; three or more medium ones give Pressure; one or two give Watch; all present gives In order. An unanswered row reads Unanswered and holds the band. The bulk "Don't know" buttons therefore never move the checklist towards In order.
Guasch's tabulated outcomes for the renegotiated concessions: delays in investment obligations in sixty-nine per cent, tariff increases in sixty-two per cent, reduced investment obligations in sixty-two per cent, changes to the asset base favourable to the operator in forty-six per cent, extensions of the term in thirty-eight per cent, and fee adjustments favourable to the operator in thirty-one per cent. The fiscal impact on the state is reported as adverse and common rather than tabulated. The levers listed against each driver are meant to give the authority a route inside the contract before that bargaining starts: the contract's own adjustment mechanisms, its security, its information rights and its dispute avoidance machinery. None of them changes an obligation or a sum; anything that would is a change under the change procedure, approved and disclosed as one.
The optional step on the Contract tab runs the 45 completeness patterns over the pasted text, with the same heading index, locator and definitions-block ranking as the other tools in this series. A mechanism found in the operative text is proposed as Present with its locator and a quote; one found only in the definitions is proposed as Present with that noted; one not found is proposed as Mark for review. Rows the profile excludes are not proposed. Every proposal needs an explicit Accept; the bulk buttons accept all of one kind at once, leave out mechanisms found in the definitions only, and say so in the export. Accepting a revenue-model proposal can change which drivers and rows apply. No accuracy figure is claimed.