Delegated but Not Discharged: Prudential Recognition and the Contingent Liability of Delegated PPP Infrastructure Guarantees in Indonesia

Abstract

Indonesia channels PPP guarantees through a state-owned guarantee enterprise to ring-fence the state balance sheet, yet in early 2024 commercial lenders still asked for the Minister of Finance’s signature, and officials acknowledged its guarantees had not substantially lowered project cost of funds. This study asks why lenders report treating an enterprise-issued guarantee less favourably than a ministerial one, why commitments to strengthen the guarantor were not expected to close that gap, and what follows for reporting the state’s contingent liability. An abductive single-case study drew on nine focus group discussions held January– March 2024 with 24 informants, read against then-current regulations and, as subsequent corroboration, Regulation 68/2024 and 2025 audited financial statements. Commercial lenders described prudential recognition, whether supervisory collateral-eligibility rules admit the instrument, as decisive for classification and, by their account, pricing; a multilateral creditor reached the same verdict through the guarantor’s status, holding that even unlimited liquidity support would not change its assessment. Guarantee scope and payment path further distinguish the instrument. Claims paid on two water projects did not transfer credibility to later instruments. Subsequent regulation consolidated co- guarantee, restoring the ministerial signature. The 2025 statements disclose guarantee exposure but no provision against Rp4.2 trillion of potential claims, about Rp1.0 trillion on PPP projects, while the central government reports no guarantee obligation. Loan spreads and expected losses are not observed. The study proposes prudential recognition as a determinant of guarantee value outside the guarantee relationship, and finds that delegation relocated recognition without relocating exposure.

Introduction

Governments in emerging economies finance infrastructure through public–private partnerships (PPP) because the budget cannot carry the whole capital programme, and a government guarantee is often what makes a PPP financeable. A guarantee is a contingent liability: it consumes no appropriation when given, appears in no expenditure line, and escapes the scrutiny a cash subsidy of the same economic value would attract. Polackova (1998) placed this class of obligation among explicit contingent liabilities, attractive to fiscal authorities because it is cheap today and potentially costly later. Heald & Hodges (2018) show that financial reporting, statistical accounting, budgeting and long-term projection characterise guarantees differently, which makes it attractive to keep them off budget and off balance sheet. Bova et al. (2019) show the cost: across eighty economies the average realisation of a contingent liability was about six per cent of GDP.

Indonesia answered this problem institutionally. Under Presidential Regulation 78 of 2010, PPP infrastructure guarantees are provided not by the Minister of Finance directly but through an infrastructure guarantee business entity, PT Penjaminan Infrastruktur Indonesia (Persero), a wholly state-owned company. The entity appraises the project, signs the guarantee agreement, pays any claim from its own balance sheet and recovers from the contracting agency under a separate recourse agreement (Presiden Republik Indonesia, 2010). The design intention is a fiscal ring-fence: an appraisal filter between political demand and the state balance sheet, and a corporate buffer that absorbs a claim before it reaches the budget. The design belongs to a wider family of state-led de-risking vehicles (Mertens & Thiemann, 2019), inside a state that has relied heavily on its own enterprises to close an acknowledged infrastructure deficit (Anguelov, 2024).

The empirical puzzle is that, at the time of our fieldwork in early 2024, the market did not treat the ring-fenced instrument as a substitute for the sovereign one. Lenders in our sessions described asking for the Minister’s signature, as a guarantee letter or as a guarantee agreement co-signed by the Ministry of Finance, and a senior official in the guarantee- policy directorate stated that government guarantees had not yet substantially affected project cost of funds (Kirichenko et al., 2025). The practice is not confined to our informants’ accounts: the Minister of Finance regulation enacted after the fieldwork, Regulation 68 of 2024, consolidates co-guarantee by the Minister and the entity as a standing option (Menteri Keuangan Republik Indonesia, 2024). By the end of 2025 the entity had guaranteed 60 projects, 51 of them infrastructure projects (PT Penjaminan Infrastruktur Indonesia, 2026). Premium, unfamiliarity and the absence of a claims record are the usual explanations.

A further observation complicates them. Asked whether an explicit government commitment to provide capacity and liquidity support sufficient to meet any claim would change the assessment, a multilateral creditor answered that it would not, because a sovereign can raise taxes and issue bonds whereas an enterprise is limited by the capital invested in it. If a binding promise of unlimited liquidity does not move the assessment, the less favourable treatment lenders describe is unlikely to be about payment capacity alone, and capacity- building on its own may not change it. One creditor’s answer cannot settle the matter, and we therefore treat capacity as a rival explanation throughout; but the answer directs attention to what else determines how the instrument is classified.

The fiscal risk matrix distinguishes direct from contingent obligations and explicit from implicit ones (Polackova, 1998). A PPP guarantee is the canonical explicit contingent liability: legally binding, triggered by an event that may not occur, and unbudgeted until it does. Implicit obligations are those the state is expected to honour without having promised to, and the expectation is created by the state’s observed pattern of rescue rather than by a signature. Delegation attempts to move a guarantee from the sovereign’s explicit contingent liabilities into the balance sheet of a separate legal person. The accounting question is whether the move is real, or whether it converts an explicit sovereign obligation into an implicit one. The largest realisations recorded by Bova et al. (2019) arise from the financial sector and from state-owned enterprises, that is, from obligations no government had signed. Bo et al. (2017) trace the same mechanism through China’s urban construction investment bonds, where debt carried by state-owned vehicles is read as sovereign long before any government recognises it, and Jin et al. (2023) show that when expectations of state support are disturbed, repricing is immediate and large.

Public sector conceptual frameworks recognise a liability when an outflow is probable and reliably measurable, and

disclose a contingent liability otherwise. In Indonesia the central government reports under the Government Accounting Standards, while the guarantee entity, as a state-owned limited liability company, reports under the Indonesian Financial Accounting Standards. Delegation through a wholly state- owned vehicle satisfies the legal form of transfer while leaving the economic substance ambiguous, because the transferee’s capacity is itself a function of the transferor’s future capital injections. Entities positioned between the state and the market account against several logics at once, which can leave them answerable to neither audience for the figure that matters (Grossi et al., 2022). A guarantee can therefore satisfy each applicable disclosure rule and still leave the reader without a measure of the state’s exposure.

Collateral-eligibility and credit risk mitigation rules determine whether it reduces the risk weight of an exposure, releases concentration limits, and permits a loan to be classified as performing with no impairment allowance. In Indonesia these rules sit in two Financial Services Authority instruments: the asset quality regulation, which classifies the portion of an earning asset secured by cash collateral as current and excludes it from the general allowance (Ketua Dewan Komisioner, 2019), and the legal lending limit regulation, which defines credit risk mitigation techniques and treats guarantees issued by the Government of the Republic of Indonesia as a distinct category (Barnes, 2024; Ketua Dewan Komisioner, 2018). Eligibility is priced: Mésonnier et al. (2022) find that loans becoming eligible as central bank collateral were funded some eight basis points more cheaply than comparable ineligible loans, and Van Bekkum et al. (2018) show that widening eligibility expands credit supply. Guarantees work through the same channel from the capital side, altering underwriting incentives (Carletti et al., 2023a). In each case value is conferred by an authority that is neither borrower nor guarantor.

Table 1 maps the studies closest to this one. Each explains part of the problem; none connects the prudential treatment of a guarantee by the beneficiary’s supervisor to the recognition and disclosure of the guaranteeing state’s contingent liability.

This study connects these literatures. It examines whether prudential recognition, that is, the collateral-eligibility rules issued by the beneficiary’s own supervisor, shapes the value of a delegated guarantee to regulated lenders alongside, and in their accounts ahead of, the guarantor’s legal standing, capital and claims record, and it traces what that displacement implies for the recognition and disclosure of the guaranteeing state’s exposure.

We use prudential recognition in a bounded sense: whether a lender’s supervisor admits a guarantee as eligible collateral or credit risk mitigation, and therefore whether the guaranteed exposure is excluded from the legal lending limit, classified as current, spared an impairment allowance and relieved of capital. It is distinct from five neighbouring properties. Legal enforceability concerns whether the guarantee is a valid obligation that a court would enforce against the guarantor; payment capacity, whether the guarantor can pay; scope, which obligations and risks are covered; fees, what the guarantee costs; and claims history, whether the guarantor has paid before. A guarantee may be enforceable, adequately capitalised, reasonably priced and

tested by paid claims and still not be recognised, because recognition turns on the supervisor’s criteria for the issuer and the instrument’s terms. The properties are not independent: scope and the route by which proceeds are paid bear on whether those criteria are met. We therefore expect prudential recognition to matter where three conditions hold: the party relying on the guarantee is prudentially regulated; the supervisor defines eligibility by issuer category and instrument terms rather than case by case; and the guarantee is meant to substitute for the borrower’s credit in the lender’s assessment. Where the relying party is not bound by such rules, as with the multilateral creditor in this study, the guarantor’s status and capacity may be assessed directly.

The study accordingly pursues three objectives: to establish, from lenders’ and creditors’ own accounts in 2024, on what grounds they report treating a guarantee issued by the guarantee business entity differently from one signed by the Minister of Finance; to explain why, in those accounts, commitments to strengthen the entity were not expected to change that treatment; and to examine how the treatment gap has since been addressed in regulation and how the relevant reporting entities recognise and disclose guarantee-related obligations and exposures in their 2025 audited statements. The study does not measure a loan-pricing discount; it reports lender-described classification and pricing.

Methods

Research Type

This study adopted an abductive qualitative single-case design, the case being Indonesia’s PPP infrastructure guarantee regime. The case is revelatory in that the research forum obtained simultaneous access to every link in the guarantee chain, from the guarantee-policy regulator to the ultimate creditor. Abduction is appropriate because the aim is to explain an observation that existing delegation theory does not anticipate rather than to confirm that theory (Timmermans & Tavory, 2012). The study tests no hypotheses; it develops propositions, stated at the close of the discussion. The unit of analysis is the guarantee instrument as valued by each actor group.

Provenance of the Data

The group discussion data were generated by a policy study of the Directorate of State Financial Risk Management,

Directorate General of Budget Financing and Risk Management, Ministry of Finance, titled Kajian Optimalisasi Peran Penjaminan Pemerintah dalam Pemenuhan Kesenjangan Pembiayaan Pembangunan Nasional (a study on optimising the role of government guarantees in closing the national development financing gap), conducted between January and March 2024 to inform policy on the governance of government guarantees. The study’s documents refer to the sessions as an interview series combining in-depth interviews and brainstorming discussions; we refer to them as focus group discussions because each session brought together several informants from one actor group. The first author led the discussions as a member of the study team, and officials of the Ministry were present in the sessions. Use of the data was authorised by the Directorate in Official Memorandum ND-321/PR.5/2024 of 17 July 2024, on condition that the names and institutions of informants be anonymised.

Temporal Structure of the Evidence

Three layers of evidence are distinguished throughout (Table 2). The focus groups took place under the rules then in force: Presidential Regulation 78/2010 and its implementing regulation, Minister of Finance Regulation 260/PMK.011/2010 as amended by 8/PMK.08/2016, together with Financial Services Authority Regulations 32/POJK.03/2018 as amended by 38/POJK.03/2019 and 40/POJK.03/2019. Minister of Finance Regulation 68/2024, enacted on 24 September 2024, revoked Regulation 260/PMK.011/2010 after the fieldwork; it is used only as subsequent corroboration of practices informants described. The audited 2025 financial statements of the central government and of the entity, published in 2026, describe the reporting position after the fieldwork and are not used to test what informants said in 2024.

Population and Sample/Informants

The population comprised the organisations that issue, regulate, receive or rely upon infrastructure guarantees in Indonesia’s PPP programme. Nine focus group discussions were conducted with twenty-four informants from seven actor groups (Table 3). Sampling was purposive and structured by position in the guarantee chain, so that each link between the issuing ministry and the ultimate creditor was represented. Institutions were invited through the policy study, and each nominated officials with direct responsibility for designing, issuing, appraising or financing guaranteed projects. Because the number and composition of sessions were fixed in advance by the policy study, saturation was not used as a stopping rule, and we do not claim saturation across the study. Each actor group met once, so adequacy was judged within each session. The moderator worked through the

themes of the terms of reference; typically after about ten to twelve questions, informants’ answers began to repeat points already made in that session, and the moderator then turned to related enrichment questions. Within-group adequacy in this sense reflects the moderator’s judgement during the session, not a count of new codes, and it cannot show that a second session with the same group would have added nothing. Commercial lenders are represented by four banks only, and statements about lender treatment are therefore statements about these four banks.

Research Location

The study was conducted in Jakarta, where the Ministry of Finance, the guarantee entity, the contracting ministries and the head offices of the participating lenders are located. Eight sessions were held in person at the offices of participating institutions and one online. Jakarta is where guarantee policy is made, instruments are negotiated and financing decisions on nationally significant PPP projects are taken.

Instrumentation or Tools

The sessions followed the discussion aims set out in the study’s terms of reference: to obtain information on the stages and practice of guarantee provision; to examine the results of government guarantees to date; to relate these to fiscal sustainability, to the financial characteristics of the guarantee instrument and to guarantee governance, from the perspectives of the government, the guarantee entity and the beneficiaries; and to capture each party’s expectations as the basis for a gap analysis and policy recommendations. The terms of reference were sent to informants before each session and are reproduced in Appendix A. There was no fixed list of questions. Sessions opened with the common themes of the terms of reference, namely the PPP guarantee arrangements, how they have been managed to date, the role of each party, and what each party had experienced and considered in need of improvement; the moderator then followed up informants’ answers. Within this common frame, lenders were probed on classification and pricing, regulators on mandate and fiscal exposure, and sponsors on bankability. The format changed in two respects during fieldwork: S8 was held as a consultative meeting between senior Ministry officials and the entity’s board, and S9 was held online in English. Because questions were not fixed, rival explanations were not put to every group in the same form; they are assessed from what informants raised (Appendix C). NVivo 14 was used to organise and retrieve coded material; no analytical claim rests on the software.

Data Collection Procedure

Sessions ran from about one to three hours, were recorded with participants’ consent and were transcribed verbatim.

Discussion was predominantly in Indonesian, with sessions S4 and S9 conducted wholly or partly in English. Quotations were translated into English by the first author and checked against the Indonesian transcripts by the second author; disagreements were resolved by returning to the recording. Documents were obtained from official gazettes and the Ministry of Finance legal database, the Financial Services Authority regulatory database, and the published audited statements (Appendix B).

Data Analysis

Coding for this article was carried out on the full set of transcripts after fieldwork had ended. It was abductive and proceeded in three levels, following Gioia et al. (2013): first- order concepts in informants’ own terms; second-order themes, formed by grouping concepts that described the same mechanism across actor groups; and aggregate dimensions, formed by grouping themes that answered the same research objective. First-order concepts were drawn from the focus groups only; documentary evidence was coded separately, is reported in Tables 5 and 6, and appears in the data structure (Figure 1) only as dated corroboration, drawn with dashed borders and not treated as informant concepts. The second author coded all transcripts; the first author reviewed every code assignment and the two resolved differences by discussion. Rather than claiming three source types for every finding, we link each claim to the informants, regulatory provisions and financial-statement notes that support it, and to contrary evidence where it exists (Table 6). Four rival explanations (generalised institutional weakness, unfamiliarity, fees, and the absence of a claims record) were specified before coding; payment capacity, guarantee scope and the payment path were added as alternatives after the first round of analysis. For each, we stated in advance of the assessment what evidence would support it and what would count against it (Appendix C, Table C1). Appendix C, Table C2 shows representative coding decisions, including cases where a segment could have been placed under more than one theme. The audit trail (memos, codebook versions and the claim-by-source matrix) is available from the corresponding author.

Result and Discussion

The coding yielded four aggregate dimensions (Figure 1): categorical non-recognition, balance-sheet finitude, payment- path and scope misalignment, and re-sovereignisation and its accounting consequence. Each dimension rests on informants’ accounts from the 2024 sessions; where later regulation or the 2025 statements corroborate them, they are identified as such in the text and drawn with dashed borders in Figure 1. Categorical Non-Recognition

Commercial lenders described classifying a government guarantee rather than evaluating it. As one lender put it in February 2024 (LEN), banks are a highly regulated industry, and what determines the treatment of a security interest is the supervisor’s rule rather than the guarantee document. Where a loan is secured by a government guarantee that is unconditional and irrevocable, the lenders described the facility as falling within the cash collateral category, with three consequences: exclusion from the legal lending limit computation, classification as current, and no impairment allowance (Ketua Dewan Komisioner, 2019, Arts. 29 and 32 and elucidation; Ketua Dewan Komisioner, 2018, Arts. 41(5) and 43). The same informant stated, unprompted, that the rate offered is far lower when such a guarantee is present, and a second lender (LEN) explained the channel: when the

capital charge falls to zero, the largest component of the price disappears. These are lenders’ descriptions of pricing, not observed spreads.

A guarantee signed only by the guarantee entity did not, in these accounts, enter that category. Lenders described such an instrument as treated like any other collateral, with no effect on the impairment allowance or the lending limit. The constraint they named was textual. A lender at a state-owned bank (Bank B in Table 4) recounted being sent the presidential regulation establishing the entity’s guarantee role and reported his compliance function’s response: unless a regulation of the Financial Services Authority says so, a presidential regulation carries no weight. The same function read the government guarantee letter and the entity’s guarantee as two different instruments.

Recognition also differed between banks. One lender (Bank A) described a consultation with the supervisor that produced an understanding that a loan carrying a guarantee given together with the Ministry of Finance falls within the cash collateral category, whatever the instrument’s form, so long as the Ministry signs; the informant qualified that this applied to that bank only, and a lender from another bank (Bank C) confirmed that his bank had reached no such understanding. Table 4 records the treatment each bank described. The evidence is limited to four banks and to their own accounts; we did not obtain supervisory correspondence. Balance-Sheet Finitude

The multilateral creditor (MDB, March 2024) described the two guarantors in binary terms. A sovereign can raise taxes, issue bonds and obtain additional financing; an enterprise is limited by the capital the government has invested in it. In this informant’s account the difference would disappear only if the guarantee were issued by the Minister, and a regulatory commitment to supply capacity and liquidity support under any condition would not make the instruments comparable in pricing. The answer places the difference in the category of guarantor rather than in the quantum of capacity behind it: a promise of unlimited liquidity support, which would answer a capacity concern, left the assessment unchanged. The multilateral creditor, which is not bound by the Indonesian supervisor’s collateral rules, thus reaches the commercial banks’ verdict by a different route, through the guarantor’s status rather than through prudential eligibility.

Table 5 sets out the figures relevant to capacity at 31 December 2025, separating nominal exposure, capital, reserves, realised claims, identified potential claims and expected loss. The entity reported equity of Rp17.49 trillion, of which paid-up share capital, representing cumulative state equity participation, was Rp10.66 trillion (PT Penjaminan Infrastruktur Indonesia, 2026, Schedule 1 and Note 19). Its annual report states guarantee exposure of Rp109 trillion on 60 guaranteed projects, 51 of them infrastructure projects under PPP and non-PPP schemes and nine non-infrastructure projects, and a gearing ratio of 6.6 times; the national economic recovery portfolio is reported separately, with exposure of Rp4.5 trillion and a gearing ratio of 2.9 times (PT Penjaminan Infrastruktur Indonesia, 2026). The audited statements do not publish the computation. As a check, exposure divided by equity excluding the Rp1.57 trillion national economic recovery injection (Rp15.91 trillion) gives 6.5 times on infrastructure exposure of Rp104 trillion and 6.9 times on total exposure, bracketing the reported ratio. Total liabilities were Rp231.36 billion, or 1.31 per cent of total assets of Rp17.72 trillion: the entity is almost entirely equity-funded (Schedule 1).

An executive of the entity (GUA, February 2024) described capacity measures in similar terms. Raising the permitted gearing ratio from ten to twelve times lifted guarantee capacity by roughly fifteen to twenty per cent but, in his words, did not

substantially improve the entity’s liquidity position. Raising the ceiling from ten to twelve times raises the maximum exposure a given equity base can support by up to 20 per cent, which is consistent with his estimate; it does not add cash.

With a reported ratio of 6.6 times at the end of 2025, the ceiling was not the binding constraint on issuance, which is itself consistent with the argument that capacity limits do not explain lenders’ treatment. He also noted that the recoverability

of recourse receivables depends on the type of contracting agency and on budgeting arrangements. The only realised claims in the regime arose on two municipal water supply projects, and the 2025 statements show gross recourse and guarantee receivables of Rp389.51 billion against which the impairment allowance rose from Rp57.69 billion in 2024 to Rp94.02 billion in 2025 (PT Penjaminan Infrastruktur Indonesia, 2026, Note 7). These figures describe the scale at which the recourse mechanism has been tested; they do not show how it would perform against a large claim.

Payment-Path and Scope Misalignment

Three features that are often conflated in discussion of the instrument can be separated. The first is scope. The entity guarantees the contracting agency’s financial obligations arising from specified infrastructure risks, principally political risk, land acquisition delay and contract termination (Presiden Republik Indonesia, 2010, Art. 4; Menteri Keuangan Republik Indonesia, 2024, Art. 74). The presidential regulation, in force at interview, confines guarantees to infrastructure risks that the contracting agency is better able to control or absorb, or that originate from the contracting agency or other parts of government. It does not guarantee the borrower’s credit. A lender (LEN) stated the distinction plainly: the entity guarantees the political issues in the project, not the bank’s credit.

The second is the payment path. The beneficiary of the guarantee is the project company, not its lenders (Menteri Keuangan Republik Indonesia, 2010, Art. 1(14), in force at interview; Menteri Keuangan Republik Indonesia, 2024, Art. 1(47)). A lender (LEN) explained that when a claim is paid, the money does not come to the bank: the bank must write to the

entity to have the claim proceeds charged in its favour, and the proceeds are remitted to the security agent’s account. That, in his account, is not cash collateral, which the bank must hold, and the arrangement is not even a pledge. An instrument whose proceeds are payable to a third-party account is therefore difficult to fit within a rule that defines eligibility by the bank’s possession of the collateral, regardless of the guarantor’s strength.

The third is timing. Under the presidential regulation in force at interview, and again under Regulation 68/2024, the entity examines a claim and pays only where it falls within the guaranteed scope and no dispute remains between the contracting agency and the beneficiary over the underlying

invoice (Presiden Republik Indonesia, 2010, Arts. 12–14; Menteri Keuangan Republik Indonesia, 2024, Arts. 80–83). Each step is defensible as stewardship of public money. Together they insert uncertainty at the point at which the beneficiary needs liquidity, and lenders said as much: once an instrument’s documentation refers to further examination, the internal reading is that it will be difficult, and the discussion stops. Because the lending limit regulation’s elucidation treats an unconditional guarantee as one free of procedural requirements that could reduce the bank’s benefit (Ketua Dewan Komisioner, 2018, elucidation to Art. 43), this feature bears directly on eligibility. We have no evidence on how long actual claims took.

Re-Sovereignisation and Its Accounting Consequence

Lenders described two routes to making the entity’s guarantee bankable for a project that would not otherwise clear credit committee. One informant (LEN, February 2024) said

there were only two: the entity issues the guarantee together with the Ministry of Finance, or the government persuades the supervisor to change its regulation so that the guarantee may be classified as cash collateral, because the bank is constrained by its regulator and knowing the shareholder changes nothing. A senior official (REG) described ring-fencing as a conflicting concept, because the party receiving the guarantee reads the state’s attempt to keep its distance as a risk in its own right, and noted that this ultimately fires back to the state budget.

Regulation enacted after the fieldwork corroborates the first route. Minister of Finance Regulation 68/2024, enacted in September 2024, provides for co-guarantee by the Minister and the entity for different risks, or for the same risk at different values, within a single PPP project, and provides that the contingency funds required are allocated through the state budget mechanism (Menteri Keuangan Republik Indonesia, 2024, Arts. 73 and 79). The regulation retains the entity as the preferred route, but consolidates the structure that restores the sovereign signature. We do not observe whether any lender’s treatment changed after the regulation. Figure 2 contrasts the three structures and the claim and recourse paths each creates.

The reporting position after the fieldwork can be read from the two sets of 2025 audited statements (Table 5), each prepared under a different framework. In the central government’s statements, government guarantees are backed by a guarantee reserve fund of Rp13.675 trillion, unchanged from the prior year and pooled across eleven guarantee programmes, of which PPP infrastructure guarantees are one (Kementerian Keuangan Republik

Indonesia, 2026, p. 161); no guarantee obligation expenditure was realised or budgeted for 2025, on the basis that no guaranteed party defaulted. The exposure delegated to the entity appears on the government’s balance sheet only as an equity investment of Rp17.49 trillion, measured by the equity method (Kementerian Keuangan Republik Indonesia, 2026). In the entity’s statements, guarantee exposure of Rp109 trillion is disclosed in the notes, and the directors record that, having reviewed potential and actual claims, no claim provision was considered necessary at 31 December 2025 or 2024 (PT Penjaminan Infrastruktur Indonesia, 2026).

The same notes disclose identified potential claims totalling approximately Rp4.2 trillion. Of this, about Rp1.0 trillion relates to PPP guarantees: Rp588 billion for land acquisition delay on six toll road projects, and Rp55 billion and Rp353 billion for a potential payment default and a potential termination on two water supply projects. The remaining Rp3,231 billion relates to a potential breach by one non-PPP debtor in 2026. These are management estimates of potential claims, disclosed but not provided for. The government’s statements report no guarantee obligation because the obligations are the entity’s; the entity’s statements report no provision because, in the directors’ judgement, the recognition threshold is not met. Neither report is incorrect on its own terms, and a nil provision is consistent both with effective risk selection and with an untested tail. What the two reports do not provide, separately or together, is an estimate of the state’s expected loss on the delegated programme; this study does not estimate one either.

Interpretation of Key Findings

The four dimensions describe a single mechanism. For the regulated lenders in our sample, the value of a state guarantee was conferred in the first instance through the beneficiary’s supervisor, by rules the guarantor neither writes nor controls; the guarantor’s standing and capacity mattered to these lenders chiefly through those rules. The eligibility effect documented in banking research arrives here through capital and classification rather than through any judgement about the guarantor’s standing (Mésonnier et al., 2022; Van Bekkum et al., 2018; Carletti et al., 2023), and the practical consequence is that a guarantee can be legally sovereign and adequately capitalised and still do little for the regulated lender whose price it was meant to move. The multilateral creditor, outside the supervisor’s rules, reached the same verdict through the guarantor’s status. Table 7 separates four components of guarantee credibility that are often treated as one.

The components interact. Scope and the payment path are not rivals to prudential eligibility; they help explain why the entity’s instrument does not satisfy the eligibility criteria even where the guarantor is well capitalised. Legal standing, in the banks’ accounts, mattered only through prudential eligibility. Payment capacity figured in the multilateral creditor’s reasoning, but as a question of the guarantor’s category rather than of its quantum; we cannot exclude that capacity also matters to commercial banks in ways they did not articulate.

Why did informants expect capacity-building not to change lenders’ treatment? The explanation we offer is recursive rather than empirical. The state’s promise to support the guarantor is itself a contingent liability of the state: unbudgeted, conditional on a future event, and enforceable only to the extent that a future budget process honours it. It therefore inherits the credibility problem it was introduced to solve, which is why a creditor that discounts the first promise can be expected to discount the second too. The reasoning mirrors the pricing of implicit sovereign support elsewhere: what a creditor buys is the probability of state rescue, read from the state’s revenue capacity and revealed behaviour rather than from the vehicle’s accounts (Bo et al., 2017; Jin et al., 2023; Chen et al., 2024). The argument is analytical; the direct evidence consistent with it is the multilateral creditor’s answer and the entity executive’s account of gearing relaxation.

The accounting consequence is the contribution of this study. The ring-fence is preserved in form and, in the market’s treatment, dissolved in effect, and the gap between the two is not represented in either set of financial statements. This resembles the pattern Heald & Hodges (2018) identify: each mode of accounting treats the guarantee correctly on its own terms, and no mode assembles the exposure. Hybridity makes the outcome stable, because the vehicle answers to a corporate audience and to a fiscal one (Grossi et al., 2022) .

Rival Explanations

Six explanations other than prudential recognition were assessed against the evidence. Several are complementary rather than competing, and none is dismissed on the strength of a single informant.

Payment capacity. Tested and not supported as the primary explanation, though not excluded. The entity’s equity is small relative to its exposure (Table 5), and the multilateral creditor reasoned from its finite balance sheet. Against a capacity explanation, the same creditor rejected unlimited liquidity support as a remedy, which a capacity concern would accept; the commercial banks described their treatment in terms of eligibility categories rather than the entity’s capital; the reported gearing ratio of 6.6 times sits well below the regulatory ceiling; and an executive described capacity

relaxation as leaving the entity’s position unchanged. Capacity may still matter at the margin, and our evidence cannot establish its weight for commercial banks.

Guarantee scope. Supported, and complementary. Because the entity guarantees the contracting agency’s obligations rather than the borrower’s credit, the instrument does not substitute for the credit assessment that drives price, whatever its eligibility.

Payment path. Supported, and complementary. Proceeds payable to the project company’s security agent make the instrument hard to classify as collateral in the bank’s possession. This helps explain non-recognition rather than displacing it.

Fees. Not supported by the evidence available, but not tested. The objection sponsors and lenders recorded was that the benefit was nil rather than that the premium was high; one sponsor stated that an investor will not pay a premium and receive nothing. We did not obtain premium data.

Claim history. Relevant, and partly refuted by the record. The instrument has been tested: claims were paid on the Bandar Lampung and Pekanbaru water supply projects and the entity performed. A lender (LEN) described deploying that history whenever auditors or investigators asked whether a government guarantee had ever been claimed successfully, and reported that the answer collapsed under examination because the instrument in question was a different guarantee. In lenders’ accounts, a successful claim record does not migrate across instruments. What remains true is that no large claim has been tested.

Generalised institutional weakness or unfamiliarity. Not supported as a general explanation. The same banks, supervised by the same authority, described accepting the

Table 7. Four Components of Guarantee Credibility

Component What moves it Evidence in this case Legal standing

Statute defining the guarantor’s status

A presidential regulation asserts the entity’s role; one bank’s compliance function declined to rely on it absent a supervisory rule Payment capacity

Capital, liquidity, gearing limits, recourse

Equity Rp15.91–17.49 trillion against exposure Rp109 trillion; gearing relaxation described as adding capacity without liquidity; recourse recoverability contingent on budgeting Legal rights: scope and payment path

Object guaranteed; beneficiary; route of proceeds

Guarantee covers agency obligations, not borrower credit; proceeds paid to the project company’s security agent; payment conditioned on examination and the absence of dispute Prudential eligibility

Collateral- eligibility and credit risk mitigation rules of the lender’s supervisor

Cash collateral treatment for direct guarantees; for co- signed guarantees at one bank only; not for entity-only guarantees at the banks in our sample Source: Authors

same state’s direct guarantee letter with no impairment allowance and full exemption from the lending limit, which makes the difference instrument-specific rather than country- specific. This internal comparison is the strongest evidence in the study. One lender attributed part of his compliance function’s resistance to unfamiliarity, so unfamiliarity may contribute at the margin.

Propositions

Three propositions follow for testing beyond this case. First, where the beneficiary of a state guarantee is a prudentially regulated financial intermediary, the guarantee’s value to it depends in the first instance on whether its supervisor’s collateral-eligibility rules admit the instrument, and the guarantor’s legal status and fiscal capacity matter to it chiefly through that admission; credibility is conferred by an actor outside the guarantee relationship. Second, strengthening the guarantor through capital injection, liquidity facilities or a relaxed gearing ratio raises capacity without altering prudential recognition, and so is unlikely on its own to change the treatment regulated lenders report, because the state’s promise to support the guarantor is itself an unbudgeted contingent liability and inherits the credibility problem it was meant to solve. Third, delegation relocates the accounting recognition of the contingent liability without relocating its economic exposure, and the state restores the instrument’s value by a means, co-guarantee, that cancels, for the projects concerned, the separation delegation was designed to achieve.

Comparison with Previous Studies

The findings are consistent with the delegation literature on where activity goes and qualify its expectations about what delegation achieves. North & Weingast (1989) and Majone (2001) hold that insulation from the principal raises the credibility of a commitment. The case suggests that this holds more readily for regulatory than for financial commitments, because insulation from political discretion can come at the price of insulation from fiscal capacity. Mertens & Thiemann (2019) describe the movement of activity off the public balance sheet without a transfer of ultimate responsibility; this case supplies the pricing mechanism their account leaves open: in lenders’ accounts, the beneficiary discounts the vehicle not because it doubts the state’s intentions but because the supervisor’s rule does not let it act on them.

The findings also qualify the PPP guarantee valuation literature, which has concentrated on instruments transferring demand or revenue risk and has valued them from the sponsor’s side (Carbonara & Pellegrino, 2018). An instrument may be valuable to a sponsor and of little financing value to a lender whose supervisor does not recognise it, so sponsor-side valuations may overstate a guarantee’s financing benefit where the debt provider is prudentially regulated. Against the institutional literature on PPP market performance (Casady, 2021), the case suggests that maturity resides in the market’s supporting rules as well as in the guarantee programme: recognition obtained bank by bank through supervisory interpretation is precisely what a guarantor cannot design for.

Limitations and Cautions

Seven limitations bound these conclusions. The study is a single case, and the mechanism should be tested where the prudential regime differs. Commercial lenders are represented by four banks and by their own accounts; we did not obtain supervisory correspondence or loan documentation. The pricing evidence is testimonial rather than transactional: no matched spreads are observed. The financial evidence postdates the focus groups, and the regulation that consolidates co-guarantee was enacted after

them, so neither can show how the informants’ own institutions responded. No large claim has tested the instrument, and the nil claim provision cannot distinguish effective risk selection from an untested tail. Sessions followed common terms of reference rather than a fixed question list, so rival explanations were not elicited uniformly across groups, and within-group adequacy rests on the moderator’s judgement in a single session. Finally, the data were generated in a policy study in which the first author participated and in which ministry officials were present, which may have shaped what lenders chose to say. Recommendations for Future Research

Assembling actual spread differentials across the three guarantee structures, from lender and guarantor records, would convert the central claim from testimony into measurement and would establish whether a discount exists and how large it is. Examining lender treatment after Regulation 68/2024 would show whether consolidated co-guarantee changed classification in practice. Comparative work could test whether guarantee vehicles in other jurisdictions converge on co-guarantee for the same reason. A further line would examine how collateral-eligibility rules are set and revised, and whether their fiscal consequences enter the deliberation.

Conclusion

This study examined whether delegating infrastructure guarantees from the Minister of Finance to a state-owned guarantee enterprise transfers the government’s contingent liability. On the evidence of this case, it transfers recognition without clearly transferring exposure. In 2024, the commercial banks in our sample described treating the entity’s instrument less favourably because their supervisor’s collateral-eligibility rules did not admit it, while a multilateral creditor reasoned from the entity’s finite balance sheet. Guarantee scope and the payment path reinforce the non-recognition. Commitments to strengthen the guarantor were not expected to change this treatment, and regulation enacted after the fieldwork consolidated the route lenders had named: co-guarantee, which restores the ministerial signature and reduces the separation delegation was designed to achieve.

The implications differ by audience. For the Ministry of Finance, the evidence suggests that capacity-building alone is unlikely to change regulated lenders’ treatment; the routes informants identified were a change in prudential rules or a ministerial co-signature, the second of which undoes the fiscal purpose of delegation, and the choice between them should be made explicitly as policy rather than project by project. For standard setters and state auditors, one option consistent with existing reporting is to present, in the fiscal risk disclosures that accompany the state budget and the central government accounts, the delegated guarantee exposure alongside the reserve fund and the entity’s identified potential claims, at nominal amounts and with any expected-loss estimate reported separately with its method. This is a disclosure proposal, not a change in recognition, and its legal basis lies in the existing fiscal risk reporting framework. For the Financial Services Authority, the treatment of the entity’s instrument is in effect a fiscal decision, since it affects how much sovereign signature the state must commit to make its infrastructure programme financeable, yet it is set within prudential regulation, outside the fiscal policy framework in which the cost of that signature is weighed (G. North & Wilson, 2020).

This study contributes to research on government guarantees and public sector accounting by identifying prudential recognition as a determinant of a delegated guarantee’s value for regulated lenders, distinct from legal standing, payment capacity and legal rights (Okoye & Gbegi,

2013). This extends credible-commitment and delegation theory from regulatory to financial commitments and establishes a boundary condition for the proposition that institutional insulation strengthens a commitment. It contributes empirically by assembling interview, regulatory and audited financial-statement evidence from every link of a national guarantee chain, a configuration rarely observable from outside the state, and it shows how a contingent liability can be disclosed in two sets of statements without either presenting the state’s exposure. Methodologically, it shows how abductive coding across actor groups can isolate an explanatory variable located outside the relationship under study. These conclusions are bounded by the single-case design, the four-bank lender sample, the testimonial character of the pricing evidence, the post-fieldwork timing of the financial evidence, and the authors’ position inside the policy process studied.

Declarations

Author Contributions

Maman Suhendra led the study and conceptualised it on the basis of his expertise in public–private partnerships and infrastructure guarantees, led all nine focus group discussions, curated the data, administered the project and prepared the original draft. Arif Nugrahanto designed the qualitative methodology, including the abductive coding protocol and the rival-explanation analysis, conducted the formal analysis with particular responsibility for the economic analysis of guarantee pricing and fiscal exposure, validated the findings, and reviewed and edited the manuscript.

Funding

Publication of this article was funded by Politeknik Keuangan Negara STAN (PKN STAN). The funder had no role in the study design, data collection, analysis, interpretation, manuscript preparation or decision to publish.

Conflict of Interest

Both authors are academic staff of Politeknik Keuangan Negara STAN, a higher education institution under the Ministry of Finance. The focus group data were generated in a policy study of the Ministry’s Directorate General of Budget Financing and Risk Management, which is among the organisations studied, and the first author was a member of the study team. The implications of this proximity are addressed in the Method and Limitations sections. The authors declare no other financial, professional or personal interest that could reasonably be perceived as influencing the

References

Anguelov, D. (2024). State-owned Enterprises and the Politics of Financializing Infrastructure Development in Indonesia: De-risking at the Limit? Development and Change, 55(3), 493–529. https://doi.org/10.1111/dech.12828

Barnes, A. (2024). Stare Decisis and Constitutional Guarantees: Commonwealth v Yunupingu. Sydney Law Review, 46(3), 347–361. https://doi.org/10.30722/slr.19761

Bo, L.,

Mear, F. C. J., &

Huang, J. (2017). New development: China’s debt transparency and the case of urban construction investment bonds. Public Money & Management, 37(3), 225–230. https://doi.org/10.1080/09540962.2017.1282254

Bova, E.,

Ruiz-

Arranz, M.,

Toscani, F. G., &

Ture, H. E. (2019). The impact of contingent liability realizations on public finances. International Tax and Public Finance, 26(2), 381–417. https://doi.org/10.1007/s10797-018-9496-1

Carbonara, N., &

Pellegrino, R. (2018). Revenue guarantee in public–private partnerships: A win–win model. Construction Management and Economics, 36(10), 584–598. https://doi.org/10.1080/01446193.2018.1467028

Carletti, E.,

Leonello, A., &

Marquez, R. (2023a). Loan guarantees, bank underwriting policies and financial stability. Journal of Financial research. Generative Artificial Intelligence Use Statement Claude (Anthropic), a large language model, was used for three limited purposes: organising the presentation of documentary and interview data, checking the consistency of the argument across sections, and drafting revisions in response to reviewer comments. Translation of informant quotations from Indonesian into English was carried out by the authors. The research idea, design, analysis, interpretation and conclusions were developed by the authors. No generative AI tool was used to generate research data, and all AI-assisted text was reviewed and revised by the authors, who remain fully responsible for the originality, accuracy, citations, analysis and conclusions of the manuscript. No generative AI tool is listed as an author. Ethics Approval and Informed Consent The data were collected in a policy study of the Ministry of Finance and were not reviewed by a research ethics committee. The authors’ institution, Politeknik Keuangan Negara STAN, did not have rules requiring ethics committee review for secondary analysis of data from a government policy study, and no such review was therefore sought. Use of the data was authorised by the data owner, the Directorate of State Financial Risk Management, Ministry of Finance (Official Memorandum ND- 321/PR.5/2024, 17 July 2024), on condition that the names and institutions of informants be anonymised. Participation was voluntary, and informants consented to the recording of sessions. To reduce the risk of re-identification, informants are identified only by actor-group codes; banks are identified by letter and reported only as state-owned or private; role descriptions are aggregated; and statements by holders of single offices are not attributed to named positions. Recordings and transcripts are held by the first author; access is restricted to the research team and may be extended to others only in limited form and with the first author’s permission. Data Availability Statement The regulations and audited financial statements analysed in this study are publicly available from official gazettes, the Financial Services Authority’s regulatory database and the published financial statements of the entities concerned. The focus group recordings and transcripts cannot be made publicly available because informants participated on condition of confidentiality and several positions in the guarantee chain are held by single officeholders who could be identified from the material. Anonymised excerpts relevant to the reported findings are available from the corresponding author upon reasonable request, subject to the consent conditions under which the data were collected. Economics, 149(2), 260–295. https://doi.org/10.1016/j.jfineco.2023.04.013

Carletti, E.,

Leonello, A., &

Marquez, R. (2023b). Loan guarantees, bank underwriting policies and financial stability. Journal of Financial Economics, 149(2), 260–295. https://doi.org/10.1016/j.jfineco.2023.04.013

Casady, C. B. (2021). Examining the institutional drivers of public-private partnership (PPP) market performance: A fuzzy set qualitative comparative analysis (fsQCA). Public Management Review, 23(7), 981– 1005. https://doi.org/10.1080/14719037.2019.1708439

Chen, R.,

Li, H.,

Tang, X.,

Jin, C.,

Zhang, S., &

Zhang, X. (2024). Impact of implicit government guarantee on the credit spread of urban construction investment bonds. Financial Innovation, 10(1), 150. https://doi.org/10.1186/s40854-024-00722-3

Gioia, D. A.,

Corley, K. G., &

Hamilton, A. L. (2013). Seeking qualitative rigor in inductive research: Notes on the Gioia methodology. Organizational Research Methods, 16(1), 15–31. https://doi.org/10.1177/1094428112452151

Grossi, G.,

Vakkuri, J., &

Sargiacomo, M. (2022). Accounting, performance and accountability challenges in hybrid organisations: A value creation perspective. Accounting, Auditing & Accountability Journal, 35(3), 577– 597. https://doi.org/10.1108/AAAJ-10-2021-5503

Heald, D., &

Hodges, R. (2018). Accounting for government guarantees: Perspectives on fiscal transparency from four modes of accounting. Accounting and Business Research, 48(7), 782–804. https://doi.org/10.1080/00014788.2018.1428525

Jin, S.,

Wang, W., &

Zhang, Z. (2023). The Real Effects of Implicit Government Guarantee: Evidence from Chinese State-Owned Enterprise Defaults. Management Science, 69(6), 3650–3674. https://doi.org/10.1287/mnsc.2022.4483 Kementerian Keuangan Republik Indonesia. (2026). Laporan Keuangan Pemerintah Pusat Tahun 2025 (Audited). Kementerian Keuangan Republik Indonesia. https://www.bpk.go.id/lkpp Ketua Dewan Komisioner. (2018). Peraturan Otoritas Jasa Keuangan Nomor 32/POJK.03/2018 tentang Batas Maksimum Pemberian Kredit dan Penyediaan Dana Besar Bagi Bank Umum. Otoritas Jasa Keuangan. https://ojk.go.id/id/regulasi/Pages/Batas-Maksimum-Pemberian- Kredit-dan-Penyediaan-Dana-Besar-Bagi-Bank-Umum.aspx Ketua Dewan Komisioner. (2019). Peraturan Otoritas Jasa Keuangan Nomor 40/POJK.03/2019 tentang Penilaian Kualitas Aset Bank Umum. Otoritas Jasa Keuangan. https://ojk.go.id/id/regulasi/Pages/Penilaian-Kualitas-Aset-Bank- Umum.aspx

Kirichenko, A.,

Kelly, L. J., &

Koskela, J. (2025). Asymptotic Guarantees for Bayesian Phylogenetic Tree Reconstruction. Journal of the American Statistical Association, 120(551), 1569–1579. https://doi.org/10.1080/01621459.2025.2485359

Majone, G. (2001). Two logics of delegation: Agency and fiduciary relations in EU governance. European Union Politics, 2(1), 103–122. https://doi.org/10.1177/1465116501002001005 Menteri Keuangan Republik Indonesia. (2010). Peraturan Menteri Keuangan Republik Indonesia Nomor 260 Tahun 2010 tentang Petunjuk Pelaksanaan Penjaminan Infrastruktur dalam Proyek Kerjasama Pemerintah dengan Badan Usaha. https://jdih.kemenkeu.go.id/dok/260-pmk-011-2010 Menteri Keuangan Republik Indonesia. (2024). Peraturan Menteri Keuangan Republik Indonesia Nomor 68 Tahun 2024 tentang Dukungan Pemerintah untuk Pembiayaan Infrastruktur melalui Skema Kerja Sama Pemerintah dan Badan Usaha dan/atau Skema Pembiayaan Lainnya. Kementerian Keuangan Republik Indonesia. https://jdih.kemenkeu.go.id/dok/pmk-68-tahun-2024/overview Appetite? Collateral Eligibility and Credit Supply. The Review of Financial Studies, 31(3), 943–979. https://doi.org/10.1093/rfs/hhx112

Mertens, D., &

Thiemann, M. (2019). Building a hidden investment state? The European Investment Bank, national development banks and European economic governance. Journal of European Public Policy, 26(1), 23–43. https://doi.org/10.1080/13501763.2017.1382556

Mésonnier, J.-S.,

O’

Donnell, C., &

Toutain, O. (2022). The Interest of Being Eligible. Journal of Money, Credit and Banking, 54(2–3), 425–458. https://doi.org/10.1111/jmcb.12851 North, D. C., & Weingast, B. R. (1989). Constitutions and Commitment: The Evolution of Institutions Governing Public Choice in Seventeenth-Century England. The Journal of Economic History, 49(4), 803–832. https://doi.org/10.1017/S0022050700009451

North, G., &

Wilson, T. (2020). Has the Australian prudential regulation authority done enough to meet its legislated objectives and prepare Australia for the next financial crisis? University of New South Wales Law Journal, 43(2), 552–580. https://doi.org/10.53637/kxis3821

Okoye, E. I., &

Gbegi, D. O. (2013). Forensic accounting: a tool for fraud detection and prevention in the public sector. (A study of selected ministries in Kogi State. International Journal of Academic Research in Business and Social Sciences, 3(3), 1–19. Polackova, H. (1998). Contingent Government Liabilities: A Hidden Risk for Fiscal Stability. In Policy Research Working Paper. World Bank. https://doi.org/10.1596/1813-9450-1989 Presiden Republik Indonesia. (2010). Peraturan Presiden Republik Indonesia Nomor 78 Tahun 2010 tentang Penjaminan Infrastruktur dalam Proyek Kerja Sama Pemerintah dengan Badan Usaha yang Dilakukan melalui Badan Usaha Penjaminan Infrastruktur. Sekretariat Negara Republik Indonesia. https://jdih.kemenkeu.go.id/dok/perpres-78-tahun-2010 PT Penjaminan Infrastruktur Indonesia. (2026). Laporan Keuangan Konsolidasian untuk Tahun yang Berakhir pada Tanggal 31 Desember 2025 (Audited). PT Penjaminan Infrastruktur Indonesia (Persero). https://www.ptpii.co.id/

Timmermans, S., &

Tavory, I. (2012). Theory Construction in Qualitative Research: From Grounded Theory to Abductive Analysis. Sociological Theory, 30(3), 167–186. https://doi.org/10.1177/0735275112457914 Van

Bekkum, S.,

Gabarro, M., &

Irani, R. M. (2018). Does a Larger Menu Increase