Experiment I: Mousetracking Study

Participants Participants were undergraduate or graduate students recruited

from subject pools at Caltech. Twenty-eight male subjects participated in Exper- iment I. Two additional subjects participated, but their data were excluded from the analysis for reasons described below.

Stimuli One hundred and twenty familiar consumer products (e.g., backpack,

watch, flash drive) were selected based on popular categories from a pilot test and an earlier study (Kang et al., 2011).5 The product images were no larger than 320×320 pixels in size. Stimulus presentation and response recording were con- trolled by MATLAB (MathWorks, Natick, MA), using the Psychophysics Toolbox extensions (Brainard, 1997; Kleiner et al., 2007; Pelli, 1997).

Experimental procedures The experiment consists of four blocks: a willingness- to-pay (WTP) reporting block, a hypothetical purchase block, a real purchase block, and a “surprise” real purchase block. Mousetracking was used only in the three purchase blocks, and participants’ viewing time was determined from mouse events (e.g., clicks on certain parts of the screen). Subjects were told that they would earn up to $50 for completing the experiment. Detailed instructions for each part were given immediately prior to that part. Therefore, participants were unaware of the existence of two real purchase blocks while they were in the hypothetical purchase block.6 7

In the WTP reporting block, subjects were shown images of the 120 consumer products, one at a time and in random order. They were asked to state a max- imum hypothetical WTP for each item, under the restriction that whatever they would buy must be for themselves (i.e., it cannot be gifted or re-sold). In each

5_{For the complete product list, see Table B.3 in Section B.3.} 6_{See Instructions in Section B.4.}

7_{We intentionally did not counter-balance the order of the hypothetical and real conditions,} following the considerations described in Kang et al. (2011) and Kang and Camerer (2013). There might be an ordering effect in which thinking about real choices first would spill over to affect hypothetical choices. On the other hand, the spill-over effect is expected to be minimal, if any, in the hypothetical-then-real order since in the real condition participants have a strong incentive to change or adjust any behavior carried over from previous hypothetical block. In addition, previous studies that used a within-subject design found no evidence for ordering (Cummings et al., 1995; Johannesson et al., 1998). Notice also that hypothetical decision followed by real decision is a natural order for forecasting purposes since, in most applications, hypothetical decision data are gathered in advance of real decisions.

Northface Campus Backpack $0 $25 $50 $30 $35.79

A

C

B

Figure 3.1: Example screens for pre-mouse/pre-eyetracking trials (A), mousetracking trials (B), and eyetracking trials (C).

trial, subjects entered an amount between $0 and $50 using a sliding scale in $1 increments (Figure 3.1A).

Upon completion of the first part, 60 out of 120 products were selected for each subject, by the computer, for presentation to the subject during the mouse- tracking blocks. More specifically, the computer ranked products in descending order of the subject’s WTPs, except for the products with WTP of $50 (to avoid the ceiling effect), and then paired up each two adjacent products (e.g., {1st, 2nd}, {3rd, 4th}, ...). Among these pairs, the 30 pairs with the highest WTP were se- lected. One product of each adjacent pair was randomly chosen and assigned to the hypothetical trials, and the other product from each pair was assigned to the real trials. This procedure ensured that the distributions of WTPs in both of hypothetical and real blocks were matched. See Figure B.1 in Section B.3.

In the hypothetical purchase block, subjects were shown a product image with an offer price, one product at a time, and asked to make a hypothetical purchase decision and respond with a Yes or No key press. Each of the 30 products selected in the aforementioned way was presented to the subject three times (for a total of 90 trials in the hypothetical block), and with a different offer price each time. The offer prices for each product were determined as follows: (i) let Pi be the

offer price for producti, WTPi be the WTP for producti, andd be a discounting

factor; (ii) sample d from the set {0.6+α, 0.9+α, 1.2+α} without replacement

for every repeat of product i, where α is a random variable from a uniform dis-

WTP reporting (120 trials) Hypothetical (90 trials) Real (90 trials) Surprise Real (90 trials) 1st 2nd 3rd 4th · · · 57th 58th 59th 60th 1st 4th · · · 58th 59th 2nd 3rd · · · 57th 60th 1st 4th · · · 58th 59th Products High WTP Low

Figure3.2: Timeline of the experiment and an illustration of allocation of products into each choice block.

the offer price by Pi = WTPi×d. This procedure ensured that there was a bal-

anced distribution of three different price levels (low, middle, and high) so as to increase statistical power and facilitate detection of a treatment effect, if any, on the purchase behavior. The jitter helps to prevent subjects from noticing any type of pricing rule.

The real purchase block was identical in structure to that of the hypothetical block except for one significant difference. Specifically, subjects were informed that one of the 90 trials would be randomly chosen at the end of the experiment, and whatever decision they had made in the chosen trial would be implemented as real, whether that purchase decision was Yes or No. Since only one trial would count as real, subjects were instructed that there was no reason for concern on their part about spreading their budget too thinly over the different items. In- stead, they were instructed to treat each trial independently of each other, as if it were the only decision in consideration. The offer prices for each product in this block were determined as in the hypothetical purchase block.

In the final “surprise” real purchase part, the same 90 item-price pairs that had been presented in the earlier hypothetical trials were shown again. This time, subjects were asked to make a real decision on these items. That is, the subjects were told that exactly one trial of the 180 real trials, including 90 from the earlier real part and 90 from this surprise real part, would be randomly selected and implemented, based on their decision made in the selected trial. This surprise real part was designed to measure switches from hypothetical to real decisions

for a matched set of items presented once in each condition; therefore, the offer prices for each product remained identical to those in the hypothetical block.

At the end of the experiment, one trial from the real or surprise real block was randomly selected. When the subject had made a purchase decision at the offered price in the selected trial, he paid the offered price out of his $50 budget from experiment participation, and the product was shipped to him and the subject received any remainder of the $50 in cash at the end of the experiment. Otherwise, if the decision in the selected trial was to not buy, the subject received the full $50 in cash and did not receive any product.

During the pre-mousetracking part, the initial location of the anchor on the WTP scale (Figure 3.1A) was randomized for each trial and recorded. These data were used as a check for subjects’ engagement in the task and possible anchoring effects. Correlations between participants’ WTP responses and anchor positions were calculated for each subject. For two subjects, the WTP reports were highly correlated with the anchor positions (p < 0.0001) and the number of trials in which |anchor−WTP| ≤ 5 was outside two standard deviations of the group average (i.e., greater than 47 trials). Therefore, these two subjects were excluded from the analysis.

Viewing time data measurement In addition to decision and response time, we

also recorded the amount of time participants spent viewing the product image and the offer price in the mousetracking trials. In each of these trials, subjects saw two gray opaque boxes on the screen behind which a product image and the offered price were hidden (Figure 3.1B). Subjects had to click and hold the left mouse-button on one of the boxes to see the information behind it (right two panels of Figure 3.1B), and they were able to see either the product or the price at any given time. The viewing times of products and prices were recorded by tracking mouse events occurring on the boxes (clicks and releases)—that is, the time elapsed from the moment when the gray box opened to the moment when it closed, and aggregated (summed) within a trial. In addition to viewing times, “latency” was defined as the time between the final box closing and when choice was entered (i.e., key press) to capture last minute computations and contempla- tions to reach a decision. There was no time limit, and subjects could spend as much time as they wished on the task. Before the start of the hypothetical pur- chase block, subjects went through five practice rounds to become familiar with the mousetracking task.

jects. That is, one group of subjects saw the screen with a product-top and price- bottom display, and the other saw the screen with a reversed (i.e., price-top and product-bottom) display.