Why free water connections can cost less per household

A connection fee can look like a straightforward way to stretch a public infrastructure budget. But if it discourages enough households from joining, the cost of a shared pump and water tower must be spread across fewer users. In a rural Bangladesh case study, that loss outweighed the money the fee would raise.

Md. Ahasan Habib and Anna Tompsett combine household willingness-to-pay data with construction costs from a piped drinking-water network to estimate which connection price minimizes public expenditure per connected household. Their answer is zero. The paper is an arXiv preprint submitted on November 5, 2026; the landing page does not indicate journal acceptance.

A real network and a price lottery

The researchers worked in two communities in Shibganj, northwestern Bangladesh, where households commonly relied on shallow tubewells. Arsenic and bacterial contamination are wider concerns for drinking water in the country. The project built one network, selected by public lottery after the valuation exercise. It draws groundwater from a deep borehole, pumps it into an elevated tank and distributes it through pipes to household taps. Three community taps also provide access to the system.

Of 385 households that agreed to a survey, 330 participated in a willingness-to-pay exercise. Each answered a sequence of questions to identify the maximum amount they would pay for a private connection, then drew a price voucher. If the price was no higher than their stated maximum and their community won the network lottery, they could use the voucher to connect at that price. This design gave respondents an incentive to state their actual valuations rather than simply guess what the researchers wanted to hear.

The median stated willingness to pay was about $6.30. A fitted demand curve implied that roughly 88% would connect at $1, falling to 12% at $20. The price maximizing connection-fee revenue was about $9.70, but the model predicted that only about one-third of participating households would connect at that price. These are estimates from reported valuations, not outcomes observed across a full experiment of uniform prices.

Sharing the fixed costs

The system's borehole, pump and storage cost about $4,795, regardless of how many households connected. The researchers estimated pipeline and tap costs at about $74 per connected household and assumed that this variable cost rose proportionally with connections. The network's theoretical capacity was about 200 households. At construction, project records counted 135 connected households and pipes laid to 176.

The researchers combined those costs with the fitted demand curve, calculating the cost per connection after deducting fee revenue at different prices. On that measure, free connections were cheapest: about $98 per household in their model. Charging deterred enough potential users to raise the fixed cost borne by each remaining connection by more than the fee offset.

For comparison, the authors estimate that a hypothetical policymaker who knew and could charge every household its exact willingness to pay could reduce net cost to about $89 per connection. That is a theoretical benchmark, not an implemented policy. In another simulation, they multiplied every participant's stated willingness to pay while holding the demand curve's shape fixed; a positive uniform price became more cost-effective only at a scale factor of roughly 3.4.

This is a particular definition of efficiency: minimizing net spending per connection, with each connection counted equally. It is not a comprehensive welfare analysis weighing different households' benefits, health outcomes or the costs of raising public funds. Nor does a zero connection fee make construction free: someone still has to finance it.

Where the evidence is uncertain

After the valuation exercise, the team capped the actual connection charge at 1,000 taka, or about $8.40, because demand was lower than expected and households raised fairness concerns about different prices. Among valuation participants in the selected community who did not draw a free voucher, actual take-up at that price was 55%, close to the 53% predicted from their stated valuations. All 20 households offered free connections took them.

The aggregate match conceals a problem: stated willingness to pay had little power to predict which individual households connected. Take-up was 53% among those who had stated a valuation below 1,000 taka and 56% among those who had stated at least that amount. The authors suggest that information about arsenic, provided after the valuation exercise, may have changed decisions; they also acknowledge that staff may have made greater efforts to persuade households with more contaminated water. Neither explanation is established causally.

The sample was modest, the communities were chosen partly for their suitability for a network, and 55 surveyed households declined the valuation exercise. The zero-price result persisted in several alternative sample analyses, but one network cannot establish the cost structure or demand response elsewhere. The model also leaves out the administrative cost of collecting fees.

The paper's theoretical model clarifies the wider lesson: free connections are most promising when shared fixed costs are substantial and demand falls sharply as a price is introduced. Those conditions alone do not prove zero is the best price globally; a larger fee might still yield a lower net cost. And the result concerns the one-time connection charge, not necessarily the price of water use. The study reports monthly operating costs of about $44.80 for the network, intended to be covered by water fees, leaving the ongoing funding question separate from the connection decision.