Tied Together: Bundling, Bargaining, and Technology Adoption

The randomized intervention changes the decision context under which dual-spouse households value household technologies.

In the individual-decision arm, the wife and husband are interviewed separately and privately. Each spouse independently submits one willingness-to-pay bid for each technology option and acts as the sole decision-maker for that recorded spouse-technology decision. Spouses cannot communicate, consult, approve, change or veto one another’s bids.

In the joint-decision arm, spouses deliberate together and submit one mutually acknowledged household bid for each technology option. The enumerator does not participate in the deliberation, average proposed amounts or select a bid on the couple’s behalf. If the spouses do not reach a mutually acknowledged bid, disagreement is recorded and no purchase is authorized for that decision.

In both arms, participants value three purchase options: a standalone improved cookstove, a standalone solar charger and a stove-charger product. The same standardized product information, bid support, BDM procedure, ownership rules and settlement procedure apply across decision-context arms. The potential voucher, any acquired technology and any residual voucher balance are household resources. Treatment assignment therefore changes decision authority and deliberation while holding ownership and experimental incentives constant.

Each spouse receives a fixed completion payment. A single binding lottery determines whether an experimental purchase decision is implemented. The selection procedure is independent of all submitted bids, and a household can receive at most one technology option.

2026-08-04

2026-09-01

Husband and Wife Individual and joint willingness to pay (WTP) for each of the three technologies (improved cookstove, solar phone charger, non-separable bundle), elicited using the BDM mechanism.

WTP is elicited via the BDM mechanism: each participant states a bid for each technology, a random price is drawn from a pre-specified uniform distribution, and the product is purchased at that price if the bid meets or exceeds it. This mechanism is incentive-compatible under standard expected utility theory, making stated bids a reliable measure of private valuations. The bargaining power index is constructed from items measuring women's decision-making authority over household expenditures, mobility, and major asset purchases, following the Abbreviated Women's Empowerment in Agriculture Index (A-WEAI; Malapit et al., 2017) and validated procedures from Nacka et al. (2024).

Secondary outcomes and mechanism measures are:

The association between women’s pre-treatment bargaining-power index and joint willingness to pay for the standalone cookstove.
The difference between the bargaining-power association for the stove-charger product and the corresponding association for the standalone cookstove.
The within-household wife–husband willingness-to-pay gap for the standalone cookstove.
The difference between the wife–husband gap for the stove-charger product and the corresponding gap for the standalone cookstove.
The within-household husband–wife willingness-to-pay premium for the standalone solar charger.
The ratio between the estimated male charger premium and the estimated female cookstove premium.

product-attractiveness ratings.
Descriptive willingness-to-pay premiums of the stove-charger product relative to each standalone product.
Descriptive purchase indicators at realized BDM prices.

The study is a 2 by 3 mixed experimental design. Decision context is randomized between households, while three technology options are valued within households.

The target population consists of dual-spouse households in eight randomly selected villages in rural northern Haiti. A complete roster of potentially eligible households will be prepared before recruitment. Eligible households will be randomly invited using an order generated by the research team.

A household is eligible if:

it contains two cohabiting adults who are married or living in a consensual union;
it currently relies on an open fire, traditional charcoal stove or another eligible traditional cooking technology;
it lacks electricity at the dwelling;
both spouses are between 18 and 65 years old;
the husband owns or is the primary user of at least one rechargeable electronic device, such as a phone or radio; and
the household does not own a solar charger, home solar system or another in-home charging source.

Both spouses first provide informed consent and complete a common baseline questionnaire in separate interview spaces. The women’s bargaining-power module is administered privately. Both spouses then receive generic BDM training and complete the same comprehension assessment. Households failing the comprehension assessment after one standardized retraining are not randomized. This determination occurs before treatment assignment and before any product-specific WTP outcome is collected.

After the baseline and comprehension records are electronically locked, 300 households will be randomly assigned in equal proportions to the individual-decision arm or joint-decision arm. Randomization will be conducted within village-session strata using computer-generated permuted blocks of four containing two assignments to each arm. Enumerators cannot access the assignment before pre-randomization records have been completed and locked.

Within each decision-context arm, product-presentation order will be independently randomized across the six possible sequences of the cookstove, solar charger and stove-charger product. Sequence randomization will use blocks of six containing each sequence exactly once. In the individual-decision arm, spouses from the same household receive the same assigned product sequence while completing the bidding tasks separately.

All three options are valued using the same incentivized BDM procedure. A potential household voucher is used for the purchase decisions, and participants may not add personal funds. The random transaction price is not revealed until all bids for the household have been recorded. A single binding lottery first determines whether any decision is implemented and then selects the technology option. In individual-decision households, one spouse is also randomly selected, and the selected spouse’s previously recorded bid for the selected technology is implemented. In joint-decision households, the couple’s joint bid is implemented. A household can receive at most one technology option.

Not available

The order of randomization will be done by the computer for individual, joint, and for order of technology presentation. However for the decision realization we will use urn to promote trust.

The unit of randomization for the decision-context treatment is the household.

Technology option is repeated within household and is not assigned as a separate between-household treatment. Product-presentation order is independently randomized at the household level within decision-context arm.

No

N/A

Under complete bidding: the individual-decision arm produces 900 spouse-product bids: 150 households times 2 spouses times 3 technology options; the joint-decision arm produces 450 couple-product bids: 150 households times 3 technology options; the experiment therefore produces up to 1,350 WTP bid-level observations. The participant recruitment target is 600 adults; the household-level randomized sample is 300.

300 households.

he fixed sample is 300 households, with 150 households assigned to each decision-context arm. Power is calculated for the sole confirmatory intersection-union hypothesis using the cookstove and stove-charger bids from wives in the individual-decision arm and couples in the joint-decision arm. The calculation assumes a marginal WTP standard deviation of 15.60 dollars, a within-decision-unit correlation of 0.60 between cookstove and stove-charger WTP, equal variances across decision contexts, and no precision gain from covariates or fixed effects. Under these assumptions, the standard error is approximately 1.80 dollars for the joint–wife-individual cookstove wedge, 1.61 dollars for the change in the wedge between the stove-charger product and cookstove, and 3.22 dollars for the sum contrast used to test non-overshooting. For one-sided tests at the 5 percent significance level and 80 percent power, the component minimum detectable effects are approximately: 4.48 dollars for the cookstove wedge; 4.01 dollars for the change in the wedge; and 8.01 dollars for the sum or non-overshooting contrast. The target pattern is a cookstove joint–wife-individual wedge of minus 10 dollars and a stove-charger wedge of minus 5 dollars. This corresponds to a 5-dollar movement toward the wife’s individual valuation and a minus 15-dollar sum contrast. Under the assumed joint distribution of the three test statistics, the intersection-union test has approximately 93 percent power. With a 10 percent reduction in the effective sample, corresponding to approximately 135 households per arm, estimated joint power remains approximately 90 percent. The confirmatory hypothesis is supported only if all three one-sided component restrictions are rejected at the 5 percent level. The calculation does not imply adequate power for every smaller economically meaningful pattern; for example, weaker baseline divergence or smaller attenuation would produce substantially lower power.