The Average Bid System in Nepal: International Experience, Implications and Lessons for Public Procurement

Legal-status note: This analysis reflects Nepal’s Public Procurement Act, 2063 as amended by the Public Procurement (Second Amendment) Act, 2083. Procurement officers and bidders should check the latest Gazette text, regulations, PPMO standard documents and project-specific bidding document before applying the rule.

Nepal has replaced automatic reliance on the lowest evaluated responsive bid for relevant competitive tenders with selection based on the responsive bid closest to the arithmetic average. The reform addresses a familiar problem: a contractor can win through a steep discount, then struggle with cash flow, quality, progress, claims and completion. But an average is not proof of a bidder’s efficient cost. International evidence—especially from Italy—shows that average-bid auctions can restrain extreme undercutting while creating a new contest in which firms try to predict or influence other bids.

What Nepal’s amended Act changes

The Second Amendment was enacted on 25 Ashadh 2083 and its commencement clause makes it effective immediately. The amended section 25 distinguishes evaluation routes. For the single-stage and two-envelope routes addressed by subsections 25(5A) and 25(7A), the public entity first applies examination, responsiveness and evaluation requirements. It then calculates the average of the bid amounts submitted by the bids remaining within that evaluation set and identifies the bid closest to that average, provided the price is not substantially above the approved cost estimate. Section 27 directs selection of the bid determined under section 25.

The statutory text does not describe a trimmed mean, random adjustment, secret coefficient or weighted inclusion of the employer’s estimate. The defensible reading is therefore the ordinary arithmetic mean of the relevant evaluated bid amounts. The approved cost estimate is a ceiling/control test; it is not stated to be another “bid” added to the numerator. If later regulations or PPMO documents prescribe operational detail, those instruments must be read with the Act rather than invented in advance.

Source	What it establishes	What it does not establish
Amended Act, sections 25 and 27	Responsive/evaluated bid closest to the average; cost-estimate control; tie rules	No trimmed-mean formula in the Act
Public Procurement Regulations	Existing procedural detail on bid examination, evaluation and cost estimates	They must not be treated as creating a formula absent from their text
PPMO standard documents/guidance	Operational instructions, forms and e-GP implementation	Guidance cannot override the Act
Media or industry commentary	Stakeholder expectations and implementation concerns	It is not binding law

The amendment also addresses extremes and ties. A bid more than 30 percent below the approved cost estimate is excluded from evaluation under the revised framework. A price substantially above the approved estimate cannot simply win because it lies near the average; the Act provides a controlled negotiation route in specified circumstances without reducing specifications or quality. If two bids are equally close on opposite sides of the average, the lower bid is selected. If bids are exactly equal at the average, selection is by lot. Qualification remains a separate requirement: mathematical proximity cannot cure missing experience, capacity, security, responsiveness or other eligibility defects.

The average-bid provisions do not convert every form of procurement into the same contest. Consultancy selection remains governed by its applicable quality/cost methods, and section 25(9) concerns a weighted technical-financial route rather than the nearest-average rule. Direct procurement, sealed quotations, special methods and procurements governed by binding development-partner arrangements must be checked under their own provisions. The tender document must identify the applicable method.

A simple example using Nepal’s rule

Assume an approved cost estimate of NPR 100 million and five bids that have passed the applicable responsiveness and qualification screens: NPR 72m, 78m, 81m, 86m and 93m. None is more than 30 percent below the estimate. Their arithmetic average is:

(72 + 78 + 81 + 86 + 93) ÷ 5 = NPR 82 million.

The NPR 81m bid is NPR 1m from the average; the NPR 86m bid is NPR 4m away. The NPR 81m bid is therefore selected, assuming every other statutory condition is met. The NPR 72m bid does not win merely because it is lowest. If bids of NPR 80m and NPR 84m were equally close to an NPR 82m average, the amended tie rule favours NPR 80m.

International experience: similar name, different rules

“Average bid” describes a family of mechanisms, not one international standard. Some systems choose the bid nearest a mean; others discard extreme discounts using an average-derived threshold; some combine a reserve price, random coefficient or technical score. Old studies confirm historical use but cannot prove that a jurisdiction still uses the same rule today.

Jurisdiction	Mechanism and period	Status/evidence	Lesson for Nepal
Italy	Trimmed-average/average-discount mechanisms were mandatory for many sub-threshold works from 1998–2006; later scope changed repeatedly	Well-documented historical and later limited use; current Italian rules use abnormally-low-bid calculations that depend on procedure and bidder numbers, not a universal nearest-mean award	Fast mechanical screening can reduce administrative discretion, but a manipulable mean can support bidding rings
Colombia	Average-related scoring appeared under Law 1150 of 2007 and implementing rules	Historical use verified; current application depends on procurement type and current decrees, so no blanket current-use claim is made	Formula changes and disclosed scoring rules affect strategy; current implementing law matters
Japan (Nagano)	Regional reform introduced an average-related mechanism in 2003	Historical prefectural example, not a statement about all current Japanese procurement	Local experiments cannot be generalized nationally; repeated markets require cartel monitoring
China (Guangdong)	Regional reforms from around 2004 used average-linked rules	Historical regional evidence; current nationwide status not established	Regional rules and market structure determine results
Peru and Taiwan	Average/reference-price mechanisms documented in comparative procurement research	Historical regulatory evidence; present universal use not verified	Do not import labels without comparing formula, exclusions and review rights
Switzerland	Comparative literature records related abnormal-bid approaches; Swiss cartel datasets mainly concern first-price tenders	No verified nationwide nearest-average award rule today	Transparent bid data can support cartel screening regardless of award formula
United States	Limited historical examples in Florida transportation and New York procurement; mainstream practice is lowest responsive/responsible bid or best value with abnormal-bid review	Not a general US system	Targeted engineer review and performance security are alternatives to automatic averages

What the Italian evidence actually shows

Italy offers the strongest warning because the rule, data and cartel cases have been studied together. Francesco Decarolis reports that average-bid auctions were extensively used after 1998. Their relevance was reduced after 2006, restricted for larger contracts during 2008–2011, and later allowed again within defined thresholds. Administrations valued the speed of a mechanical rule because case-by-case abnormal-bid screening demands engineers, records and legal capacity.

The same design linked every bidder’s winning probability to other submitted bids. In Turin, coordinated groups could place multiple bids to move or surround the trimmed average. A court case involved 95 firms operating through cartels. In Decarolis’s empirical comparison, average-bid auctions attracted far more bids, but many were consistent with coordinated or strategically placed submissions. His counterfactual estimates found an average-production-cost inefficiency of about 16 percent and a higher-cost winner than the corresponding first-price auction in roughly two-thirds of cases. These are findings from a specific Italian dataset, not a forecast that Nepal will reproduce the same percentages.

The practical conclusion is balanced. Lowest-price procurement can select a fragile bid and generate costly adaptation. Average bidding can reduce the reward for a single extreme discount and reduce the workload of subjective rejection. Yet it weakens the link between a firm’s own cost and its bid. Once the target is an expected mean, a rational bidder asks “What will others quote?” as well as “What can I deliver for?” A cartel or one bidder controlling related entities can gain an even stronger advantage.

Potential benefits—and their limits

• Less extreme underbidding: the lowest number no longer automatically wins, and the 30-percent boundary removes the deepest discounts.
• More viable initial cash flow: a price nearer the market cluster may leave more room for labour, materials, supervision and financing.
• Fewer mechanically induced disputes: fewer bids may depend on later variations or claims to recover an unrealistic tender price.
• Administrative consistency: a stated formula can reduce ad hoc judgement if data and evaluation steps are auditable.

None of these benefits is automatic. A price near the average can still be technically weak, financially unsustainable or above efficient cost. Delays also arise from poor design, land and utility clearance, delayed decisions, late payment, weak supervision and unrealistic schedules. An award formula cannot repair deficient project preparation.

Risks Nepal must manage

• Strategic guessing: bids converge around an anticipated average instead of independent cost.
• Artificial entries: non-serious or related bidders can submit numbers intended to move the mean.
• Bid rotation and rings: repeated local markets make coordination easier, especially where firms know likely participants.
• Reduced savings: excluding vigorous low bids may raise award prices even when a low bidder could perform.
• Estimate dependence: an outdated or inflated approved estimate weakens the 30-percent screen and the affordability test.
• False confidence: “closest” is mathematically precise but does not establish value for money, quality or delivery capacity.

Nepal-specific implications

Nepal’s construction market combines chronic low bidding with delayed site possession, payment constraints, price volatility, capacity gaps and frequent time extensions. Average bidding may improve the starting price on some works, but public entities must still issue buildable designs, realistic quantities, current rate analysis, clear risk allocation and funded programmes. Engineers should document the estimate’s base date, sources, productivity assumptions, haul distances, taxes, escalation treatment and site constraints. Otherwise the mean merely clusters bidders around a weak benchmark.

Small and medium contractors face a mixed effect. They may benefit when aggressive discounts by cash-rich or overextended competitors no longer dominate. They may also lose if success depends on forecasting a market average or joining informal information networks. Open participation should therefore be paired with beneficial-ownership checks, related-party detection and scrutiny of repeated bidder groups.

Nepal’s e-GP system can make the reform auditable. PPMO should publish machine-readable bid distributions, bidder identities, ownership links where legally available, evaluation exclusions, award distance from the mean, contract progress, variations and completion outcomes. Red flags include clusters of identical discounts, firms that bid together repeatedly, losing bids that systematically position a designated winner, sudden entry of firms with no realistic capacity, and rotation across neighbouring contracts. Such patterns justify review; they are not proof by themselves.

An implementation control sequence

Approved design and current cost estimate → clear bidding document and applicable evaluation route → eligibility and responsiveness review → removal of legally excluded bids → arithmetic-average calculation with an audit trail → closest-bid and tie-rule check → qualification confirmation → collusion/red-flag screening → notice of intention to award → performance monitoring against the original bid.

PPMO and oversight bodies should compare outcomes before and after reform: number of bidders, award-to-estimate ratio, concentration, time to award, completion time, variations, claims, termination, defects and final cost. The question is not whether award prices rise or fall alone. It is whether completed assets deliver better life-cycle value.

Does FIDIC endorse average bidding?

FIDIC does not specifically endorse or prohibit Nepal’s average-bid formula. FIDIC Conditions of Contract allocate contractual rights and risks after a procurement strategy and contract form are chosen; they are not national tender-award legislation. FIDIC’s 2017 contractor-selection guideline criticises routine low-bid selection for failing to provide ultimate best value and presents alternatives such as prequalification, lowest workable cost, cost-and-quality and quality approaches. Its design-build policy likewise says selection should be based on value rather than apparent least cost.

That direction supports stronger qualification, workable pricing and whole-life judgement, but it is not approval of a nearest-average algorithm. National law governs Nepal’s award rule; donor procurement rules may govern where an agreement makes them applicable; the selected FIDIC contract then governs contract administration. These layers should not be conflated. Modern multilateral practice increasingly permits rated non-price criteria where they are proportionate, disclosed and measurable, while retaining price, qualification and abnormally-low-bid safeguards.

Will the reform solve low bidding?

It can reduce one incentive: winning by being the deepest responsive discount. It cannot by itself establish fair cost, capable management or clean competition. Without accurate estimates, ownership transparency, bid-pattern analysis and contract-performance data, Nepal could replace price undercutting with competition to predict—or manipulate—the average.

The sound test is empirical. If the reform produces fewer failed contracts, less opportunistic variation, better quality and timely completion without abnormal price convergence or concentration, it will have improved value for money. If bids cluster mechanically, related bidders proliferate and the same groups rotate awards, the formula will require correction. Public procurement should reward deliverable value, not simply the lowest number or the luckiest guess.

Related reading: Why Public Entity Estimates Are High While Contractors Bid Low in Nepal.

Sources and further reading

• Ministry of Law, Justice and Parliamentary Affairs: Public Procurement (Second Amendment) Ordinance/legislative records
• Federal Parliament: Public Procurement (Second Amendment) Bill, 2083
• Nepal Law Commission: Public Procurement Act, 2063
• Nepal Law Commission: Public Procurement Regulations, 2064
• Department of Roads: application of amended section 25(7A)
• FIDIC Guidelines for the Selection of Contractors (2017)
• FIDIC: The Role of the Consulting Engineer in Design-Build
• Francesco Decarolis, Comparing Public Procurement Auctions
• Wallimann et al., Collusion detection in public procurement auctions
• OECD: Collusion and Corruption in Public Procurement