The research, published in Frontiers in Behavioral Economics, draws on two incentivized experiments led by Christopher Gustafson of the University of Nebraska-Lincoln, Syed Imran Ali Meerza of Arkansas Tech University, and Khondoker Mottaleb of Texas Tech University. One experiment examined how negative information, an article about economically motivated mislabeling fraud in the Italian extra virgin olive oil industry, affected bids on olive oils from Italy, Greece, and the United States. The other examined how positive information, an article describing the health benefits of zinc-biofortified rice, affected bids on that rice and three common non-biofortified varieties in Bangladesh. In both studies, the researchers compared valuations elicited under the two design structures and found no statistically significant differences for the products directly discussed in the information texts.

The stakes of this design question are considerable. Within-subject designs, in which each participant provides valuations at multiple stages, generate more observations per person, which increases statistical power and allows researchers to track how individuals actually update their valuations when they receive new information. They also eliminate the risk that random assignment to treatment and control groups fails, leaving the groups unbalanced on hard-to-observe traits such as tastes and preferences. Between-subject designs, in which each participant experiences only one condition, have long been viewed as the more conservative choice, precisely because they avoid some of the psychological pitfalls that repeated measurement may introduce.

Those pitfalls are not hypothetical. In a within-subject experiment, a participant’s first bid on a product can serve as an anchor, and the well-documented anchoring and adjustment heuristic suggests later bids may be dragged toward that initial number rather than fully reflecting the new information. Repeated exposure to experiment stimuli within a single session may also allow participants to deduce what the researchers are studying, creating experimenter demand effects in which people respond the way they think the researchers want rather than the way they would in a market. Loss aversion adds another wrinkle: people tend to react more strongly to negative information than to positive information, so the two valence conditions could interact differently with anchoring. The researchers formalized these concerns in two hypotheses predicting that both the direct effects of information on implicated products and the indirect spillover effects on unimplicated substitutes would be significantly smaller in within-subject designs.

To test these hypotheses cleanly, both experiments used the Becker-DeGroot-Marschak mechanism, or BDM, an incentive-compatible bidding procedure. Participants submit a maximum bid for each product, and a price is then drawn at random from a uniform distribution. If the bid meets or exceeds the drawn price, the participant buys the product at that randomly determined price. Because what a participant pays does not depend on their own bid, truthful bidding is in their best interest. The BDM also avoids bid affiliation, the phenomenon in which participants’ bids are influenced by the bids of others, which can arise when the market price is set from one participant’s bid and which could have confounded estimates of the response to information.

The structural logic of the experiments was identical across the two studies. Participants were randomly assigned to one of two groups. Group A, the within-subject condition, submitted a round of valuations before receiving any information and a second round after reading it. Group B, the between-subject condition, read the information first and then submitted a single round of valuations. Because both groups’ valuations can be compared against the same randomized no-information baseline from Group A’s first stage, the design allows a direct statistical comparison of how much information moved willingness to pay under each design. Study 1 involved 177 regular olive oil consumers recruited at a university experimental economics laboratory in the midwestern United States between September and November 2017, bidding on 500-milliliter bottles of extra virgin olive oil from Italy, Greece, and California, with shelf prices between $7.99 and $8.99. Study 2 involved 267 adult primary rice shoppers recruited across four districts of Bangladesh in January 2022, bidding on one-kilogram samples of zinc-biofortified BRRI dhan74 rice and the common miniket, paijam, and katari varieties.

The negative information in Study 1 moved valuations dramatically, and in nearly identical ways under both designs. Without information, participants valued the Italian olive oil most highly at $7.53 per bottle, followed by Greek oil at $6.99 and California oil at $5.96. After reading about mislabeling fraud in the Italian industry, participants in the between-subject condition valued Italian oil at just $3.99, a drop of $3.53 or roughly 47 percent, with smaller significant declines for the Greek and US oils. In the within-subject condition, the Italian oil fell from $7.53 to $3.93, a decline of $3.60 or about 48 percent, again with smaller significant declines for the substitutes. Regression analysis comparing post-information valuations across the two designs found the difference for the directly implicated Italian oil was only $0.24 and statistically insignificant. Linear-combination tests for each product showed no design effect significant at the 5 percent level, although the estimate for Greek oil, at negative $0.90, came close with a p-value of 0.059.

The positive information in Study 2 told a similar story. In the between-subject condition, participants who read about the health benefits of zinc valued the biofortified rice at 47.6 Bangladeshi Taka compared with 42.5 Taka for those who received no information, a significant increase of about 11.9 percent, while valuations of the other rice varieties changed little. In the within-subject condition, bids on the zinc rice rose from 42.5 to 49.1 Taka, an increase of more than 15 percent, while bids on the three common varieties each declined by roughly 2 to 3 percent. Yet when the researchers compared post-information valuations across designs, the difference for the zinc-biofortified rice was a negligible negative 1.23 Taka, about two US cents, and statistically insignificant. None of the product-specific design effects for the substitute varieties reached significance either. Notably, the within-subject design actually produced somewhat more conservative estimates of spillover effects on the non-target products, the opposite of what the anchoring hypothesis predicted.

The authors caution that these findings come with boundaries. The two studies were conducted in different countries with different products, so they do not directly test whether responses to positive versus negative information differ; the valence contrast is exploratory. Nonsignificant differences also do not guarantee that the two designs are always equivalent. And the estimates of spillover effects on unimplicated substitutes showed greater variation across designs than the estimates for target products, so researchers whose primary interest lies in those indirect effects may want to give the design choice additional thought. The authors also point to features of their studies that may explain why they found no design differences where earlier work, including an unpublished experiment on package sizes that found a valuation premium 2.5 times larger in a within-subject design, did find differences. Their participants were highly familiar with the products they bid on, having been recruited precisely because they regularly consumed olive oil or rice, and the object of interest was a reaction to clearly attributed information rather than the evaluation of unfamiliar or hard-to-assess attributes. Familiarity is known to stabilize valuations and reduce behavioral anomalies, and evaluability theory suggests that judgments about easily evaluated attributes, such as country of origin or zinc content, are less susceptible to context effects than judgments about obscure ones.

For the broader research community, the practical implication is that the statistical advantages of within-subject designs may be available without the feared distortion, at least for incentivized valuation of familiar consumer products. Within-subject designs allow researchers to study how individual consumers update their willingness to pay when information arrives, to examine heterogeneity in that updating across people, and to avoid the randomization failures that can undermine smaller between-subject studies. The authors suggest that the choice between designs may ultimately rest on the researcher’s judgment about external validity, budget constraints, and whether individual-level change is itself the object of study. But for the common task of measuring how information about a product, whether a warning about fraud or a promise of better nutrition, shifts what consumers are willing to pay for it, the evidence from these two experiments indicates that either design will tell much the same story.

Subject of Research: The effect of between-subject versus within-subject experimental designs on consumer willingness-to-pay estimates under positive and negative information

Article Title: Estimating the impact of between- and within-subject designs on consumer valuation experiments: the case of positive vs. negative information

Article References: Gustafson, C. R., Meerza, S. I. A., & Mottaleb, K. A. (2026). Estimating the impact of between- and within-subject designs on consumer valuation experiments: the case of positive vs. negative information. Frontiers in Behavioral Economics, 5, 1870245. https://doi.org/10.3389/frbhe.2026.1870245

Image Credits: AI Generated

DOI: 10.3389/frbhe.2026.1870245

Keywords: consumer valuation, willingness to pay, within-subject design, between-subject design, BDM mechanism, anchoring, olive oil fraud, zinc-biofortified rice, information effects, experimental economics, loss aversion, spillover effects