Replication: The unit effect

Replication
Science
Decision Making
Author

MSM

Published

September 10, 2026

Original work

Pandelaere, M., Briers, B., & Lembregts, C. (2011). How to make a 29% increase look bigger: The unit effect in option comparisons. Journal of Consumer Research, 38(2), 308-322.

The unit effect: the same difference looks bigger when it is expressed in a unit that produces larger numbers. A Snickers against a mandarin is 488 versus 53 kcal, or 2042 versus 222 kJ. It’s the same gap, but 1820 feels like a lot more than 435. We ran two studies, S3 and S4, and in each one the unit was varied between participants.

Team BA spring 2026

  • Bardusha Sarveswaran
  • Benno Six
  • Celestine Jaloux
  • Celine Amherd
  • Jorden Lindner
  • Karar Al-Ubaidy
  • Lena Kowal
  • Michelle Zurwerra

Sample

A total of 196 people took part in the study. The following filters were applied:

  1. Under 18 years of age: 0 participants removed
  2. Study not completed: 37 removed
  3. Study completed too quickly or too slowly: 62 removed
  4. Already knew the study (answered ‘yes’): 39 removed
  5. At least 2 of 3 control questions correct: 46 removed

That leaves 113 in the sample. Note: the criteria are listed as applied independently, not cumulatively.

Of these, 67 are female (M = 31.3, SD = 14.6 years), 45 male (M = 31.8, SD = 13.9 years), 1 non-binary (M = 24, SD = NA years) and 0 did not report a gender.

Replication Study S3

Participants made seven choices between a lower- and a higher-calorie snack, for example a mandarin (53 kcal) or a Snickers (488 kcal). One group saw the energy content in kcal, the other in kJ. For each choice we take the energy value of the chosen snack and convert kJ to kcal (1 kcal = 4.184 kJ), so both groups are on the same scale. If the unit effect holds, the larger kJ numbers make the gap between the snacks look bigger, and the kJ group should end up with fewer calories.

The figure shows the mean over all choices in each group, with 95% confidence intervals.

Figure 1

A t-test between the two conditions (kcal v. kJ) gives:

Effect sizes were labelled following Cohen's (1988) recommendations.

The Welch Two Sample t-test testing the difference of choice_value by unit
(mean in group kcal = 308.43, mean in group kJ = 262.11) suggests that the
effect is positive, statistically significant, and very small (difference =
46.32, 95% CI [11.48, 81.16], t(784.50) = 2.61, p = 0.009; Cohen's d = 0.19,
95% CI [0.05, 0.33])

The kJ group chose lower-calorie snacks. The effect is small, but it goes in the direction the unit effect predicts.

Replication Study S4

Participants made ten choices between standard delivery and a more expensive express delivery, for example standard delivery in 62 days for CHF 499 or express 31 days earlier for CHF 534. One group saw the times in days (ET), the other in months (EM). Here the unit effect predicts that 31 days sounds like more time saved than 1 month, so the days group should pay for express more often. Two more groups saw the same choices in an attention version, in days (AET) and in months (AEM).

For each choice we take the number of days in the chosen answer: the delivery time for standard, the time saved for express. Months are converted to days (1 month = 31 days, as in the materials). Standard always carries the larger number, so a higher mean means more people chose standard delivery.

Figure 2

A t-test between the two conditions (days v. months) gives:

Effect sizes were labelled following Cohen's (1988) recommendations.

The Welch Two Sample t-test testing the difference of days by unit (mean in
group EM = 85.00, mean in group ET = 85.06) suggests that the effect is
negative, statistically not significant, and very small (difference = -0.06,
95% CI [-8.49, 8.37], t(491.91) = -0.01, p = 0.989; Cohen's d = -1.28e-03, 95%
CI [-0.18, 0.18])

The two groups chose almost identically, so for delivery times the unit effect does not replicate.

A t-test between the two conditions (months v. attention months) gives:

Effect sizes were labelled following Cohen's (1988) recommendations.

The Welch Two Sample t-test testing the difference of days by unit (mean in
group AEM = 78.18, mean in group EM = 85.00) suggests that the effect is
negative, statistically not significant, and very small (difference = -6.82,
95% CI [-14.70, 1.06], t(516.94) = -1.70, p = 0.090; Cohen's d = -0.15, 95% CI
[-0.32, 0.02])

The attention version of the months condition does not differ reliably from the plain one either.

Data

The anonymised responses behind every figure and table on this page are in BAFS2026_public.csv: one row per participant, with the random ID, the consent and job items, the answers to both replication studies (S3_* and S4_*), the Manipulation and Kontrollfrage checks, Interesse and Feedback, and the questionnaire metadata. Gender, birth year and age are held back and are not published.