gunadi_2022_impact
Six preregistered studies (N = 5,713) test how the direction and frequency of historical price changes affect consumers' decision to defer a purchase (buy now vs. later). Study 1a (N=802) and Study 1b (N=800) use between-subjects 2(direction) x 2(frequency) flight-price scenarios where participants choose buy now vs. defer (Study 1b is a sequential-page design where deferral is recorded page by page until buy-now or the final page). Study 2 (N=802) adds a belief/agreement and a price-expectation measure to test the future-price-expectation mechanism; Study 3 (N=1506) and Study 4 (N=1506) test moderators (monotonicity and timing of the price change), each with a choice plus price-prediction measure. Study 5 (N=297, business-school students) uses actual purchase decisions for 4 products, each with a free-text rationale and follow-up ratings. The core research question is whether prior price decreases (vs. increases) and more frequent price changes raise the likelihood of deferring purchase, mediated by expectations that prices will continue to drop. Response types are binary buy-now/defer choices, rated expectations, open-ended price predictions, and free text. All headline effects (price-decrease-direction effect, frequency moderation, and the consequential-choice effect in Study 5) reproduce in the local data (all p < 1e-11).
Responses
9,114
162 held out
CORE effects
2 / 2
CORE test NLL (avg)
0.515
rank 4 of 7 models
Best model
CORE (no forgetting)
test NLL (avg) 0.500
Likelihood of human responses
| model | PNLL | test NLL (avg) | test responses | train NLL (avg) | train responses |
|---|---|---|---|---|---|
| CORE | 83.5 | 0.5153 | 162 | 0.4818 | 8,952 |
| CORE (no prediction errors) | 82.9 | 0.5115 | 162 | 0.4714 | 8,952 |
| CORE (no phonemes) | 83.0 | 0.5125 | 162 | 0.4673 | 8,952 |
| CORE (no forgetting) best | 81.0 | 0.5002 | 162 | 0.4865 | 8,952 |
| CORE (no hierarchy) | 96.9 | 0.5984 | 162 | 0.6072 | 8,952 |
| Transformer | 90.8 | 0.5608 | 162 | 0.5246 | 8,952 |
| Centaur | 98.3 | 0.6068 | 162 | – | – |
Experiments
| exp | task class | participants | responses | conditions | simulated | effects |
|---|---|---|---|---|---|---|
| exp0 | FlightDeferral | 802 | 802 | 1 | yes | direction_effect_single, frequency_moderation_direction |
| exp1 | SequentialTVDeferral | 800 | 2,487 | 1 | no | – |
| exp2 | SpeakerDeferral | 802 | 802 | 1 | no | – |
| exp3 | SpeakerMonotonicDeferral | 1,506 | 1,506 | 1 | no | – |
| exp4 | TabletDeferral | 1,506 | 1,506 | 1 | no | – |
| exp5 | StoreDeferral | 297 | 2,011 | 1 | no | – |
Behavioral effects
| effect | experiment | paper | CORE | CORE (no prediction errors) | CORE (no phonemes) | CORE (no forgetting) | CORE (no hierarchy) |
|---|---|---|---|---|---|---|---|
direction_effect_single | exp0 | +0.442 | ✓ yes +0.487 | ✓ yes +0.442 | ✓ yes +0.491 | ✓ yes +0.255 | ✓ yes +0.099 |
on a single historical price change, deferral (P(buy later)) is higher after a price increase than after a decrease (chi-square). | |||||||
frequency_moderation_direction | exp0 | -2.900 | ✓ yes -2.028 | ✓ yes -2.274 | ✓ yes -2.641 | ✓ yes -0.879 | ✗ no -0.214 |
the increase-minus-decrease gap in deferral is larger for a single change than for multiple changes (logistic interaction). | |||||||
Transcript
You are shopping for a flight ticket for your upcoming holidays. Only one airline flies directly to your destination, so you must fly with it. The airline shows you the ticket's historical prices: $200 (4 weeks ago), $250 (3 weeks ago), $300 (2 weeks ago), $350 (1 week ago). Currently, the ticket costs $400. On each decision you press N if you would buy the ticket now, or W if you would wait to buy it later (defer the purchase). You decide whether to buy the $400 ticket now or wait. You press N.
Prices of products change frequently over time. You are considering buying a new TV that you have been checking the price of over the past days. You will be told the price of the TV at different points in time, page by page. On each page you press N to buy the TV now (which ends the study for you), or W to wait (buy later) and see the next price update. On January 16, 2021 the price of the TV was $1200. Currently the price is $1200. You press N. The study ends because you buy now.
You are considering buying a Bluetooth speaker that you have been checking the price of over the past weeks. The speaker's historical price was: $20 (4 weeks ago), $30 (3 weeks ago), $40 (2 weeks ago), $50 (1 week ago). Currently, the speaker's price is $60. On each decision, press N to buy the speaker now, or W to wait (buy later). You decide whether to buy the speaker now or later. You press W. Rate your agreement that 'continue to increase in the future' on a scale from 1 (strongly disagree) to 7 (strongly agree). You answer 2. Estimate the price of the speaker in a week from now (in $, between $0 and $150). You answer 50.
You are considering buying a Bluetooth speaker that you have been checking the price of over the past weeks. The speaker's historical price was: $100 (4 weeks ago), $90 (3 weeks ago), $80 (2 weeks ago), $70 (1 week ago). Currently, the speaker's price is $60. On each decision, press N to buy the speaker now, or W to wait (buy later). You decide whether to buy the speaker now or later. You press W. You predict the price of the speaker in a week from now (in $, between $0 and $150). You answer 50.
You are considering buying a tablet. The tablet's historical price was: $399 (5 weeks ago), $399 (4 weeks ago), $299 (3 weeks ago), $299 (2 weeks ago), $299 (1 week ago). Currently, the tablet's price is $299. On each decision, press N to buy the tablet now, or W to wait (buy later). You decide whether to buy the tablet now or later. You press N. You predict the price of the tablet in a week from now (in $, between $0 and $500). You answer 299.
You are testing a new online platform for the university store, which sells university-branded products and uses dynamic pricing (like Amazon), so prices may change frequently. You will review 4 products. On each product page, you decide whether to buy now or wait. If you wait, you may receive a later price update. Your decisions are consequential: after the study a random draw selects 5 participants, and your choices are carried out with €20 funded by the study. On each product, press N to buy now, or W to wait to buy. If you wait on a product, you are later asked whether you want to receive a price update by email: press Y for yes or N for no. Product: Aluminum Card Holder with Leather Case. Maximum capacity of 8 standard-sized cards. Made with soft and durable leather. Minimalist and slim design, extremely light and convenient to carry. Price history: Was €24 (4 days ago) | Was €24 (3 days ago) | Was €24 (2 days ago) | Was €24 (1 day ago). CURRENT PRICE: €15. What will you do? You press N. Product: Portable Battery Charger. External portable battery charger with 2,200 mAh capacity. Very light, aluminum finished body, with an extra-flat design. Has both MicroUSB and USB-C inputs. Price history: Was €7 (4 days ago) | Was €9 (3 days ago) | Was €10 (2 days ago) | Was €13 (1 day ago). CURRENT PRICE: €15. What will you do? You press W. You are asked whether you want to receive a price update for the battery charger by email. You press Y. Product: Reusable Water Bottle. A university-branded product. Price history: Was €16 (4 days ago) | Was €15 (3 days ago) | Was €13 (2 days ago) | Was €12 (1 day ago). CURRENT PRICE: €10. What will you do? You press W. You are asked whether you want to receive a price update for the water bottle by email. You press N. Product: Laptop Bag with Adjustable Straps. This light laptop bag comes with an adjustable strap and two front pockets. Inside, it has a padded compartment for a laptop up to 13.3 inches in size. Price history: Was €10 (4 days ago) | Was €12 (3 days ago) | Was €15 (2 days ago) | Was €18 (1 day ago). CURRENT PRICE: €20. What will you do? You press W. You are asked whether you want to receive a price update for the laptop bag by email. You press N. You briefly explain what motivated your decisions: The cardholder motivated me because there was not any price change when looking at the past days and know there was. The rest of the products, I already have or do not like.. 'How much did you like the look of the pages?' (1 = not at all, 7 = very much). You rate 2. 'How likely would you be to purchase there?' (1 = not at all, 7 = very much). You rate 3. 'How likely would you be to recommend the platform?' (1 = not at all, 7 = very much). You rate 1. 'Does the look fit the brand/image?' (1 = not at all, 7 = very much). You rate 2.