Concentration Bias in Intertemporal Choice: Evidence from Choices Involving Real-Effort and Restaurant Voucher

We conduct a laboratory experiment to study the effect of concentration bias on intertemporal choice at the CLER of the University of Cologne and at the BonnEconLab of the University of Bonn.

The experimental design and analyses described here extend the experimental design and analyses of the trial “Concentration Bias and Intertemporal Choice” https://www.socialscienceregistry.org/trials/4341/history/49221 that was conducted and completed by the end of June 2019.

The design and analyses of this trial differ from the previous trial in three ways:
1. While subjects traded off exerting more effort for generating a greater donation in the previous trail, subjects now trade off exerting effort for generating a greater value of a restaurant voucher. This allows to show robustness of the evidence provided in the previous trial.
2. We add two treatments to study the mechanism behind concentration bias. In particular, we study the role of cognitive complexity in concentration bias by displaying dispersed consequences in a cognitively easier way and by displaying concentrated consequences in a cognitively more difficult way. Finding evidence for concentration bias in these new treatments would indicate that concentration bias is not merely driven by cognitive complexity.
3. We include new measures to extend our heterogeneity analyses and remove one previous measure.

The main feature of our experimental setup are multiple pairwise intertemporal choices that participants make after reading computerized instructions and completing practice choices. In each intertemporal choice, subjects choose between two alternatives, Alternative A and Alternative B. Each alternative consists of two components:
• a real-effort work schedule that distributes real-effort tasks over up to 8 work periods in the future and is completed online by subjects using their computers at home;
• a remuneration to be paid after subjects successfully completed their work schedule; the remuneration consists of up to two parts: a restaurant voucher valid on a particular date in the future and a monetary payment.

In all choices, the remuneration is concentrated in a single period. In some choices, the work schedule is concentrated in one period and in others, the work schedule is dispersed over multiple periods. When designing choices such that standard discounting models assume that individuals should behave equally, we test whether subjects are willing to work more when work is dispersed over multiple days than when it is concentrated in a single period.

2019-07-15

2019-08-16

The study includes a primary outcome variable: subjects’ indifference points in Part 2.

See experimental design

Subjects’ indifference points in Part 1.

See experimental design

The experiment consists of three parts.
Part 1
Part 1 consists of 8 blocks of intertemporal choices. Each block consists of a sequence of 126 fixed intertemporal choices. Subjects complete the blocks in random order.
The contents and goal of the intertemporal choices within a block are as follows:
• Both alternatives, A and B, consist of work schedules that distribute positive numbers of real-effort tasks over all 8 work periods. Upon completion of the work schedule, according to both alternatives, a payment of 100 euros is transferred to subjects’ bank accounts and subjects receive a restaurant voucher.
• The alternatives differ in the value of the restaurant voucher and in the number of real-effort tasks in one of the work periods. This period is henceforth called the block-defining work period.
• The restaurant voucher is greater in A than in B for all decisions.
• The work schedule of Alternative A implies a greater number of real-effort tasks for the block-defining work period than the work schedule of Alternative B in all but one choice—in which they are the same. The work load in all 7 remaining work periods is the same in A and B.
• Alternative B is constant among all choices.
• Alternative A varies between the choices of a block only in the number of real-effort tasks in the block-defining work period.
• The intertemporal choices allow us to elicit subjects’ indifference points such that we know how many more real-effort tasks they are willing to complete in the block-defining work period in exchange for a greater value of the restaurant voucher.
• Eliciting indifference points for each block is our primary goal of Part 1.
• Part 2 will directly depend on the indifference points elicited in Part 1. Subjects are unaware that their Part-1 choices affect their Part-2 choices to keep subjects properly incentivized to report their truthful indifference points in Part 1.
Blocks
• Between the 8 blocks, we vary the block-defining work period among all 8 work periods and the voucher values in the following way: Sorting the voucher values among the blocks, the voucher value of Alternative A in a given decision block equals the voucher value of Alternative B in the preceding decision block.
Part 2
Part 2 consists of one block of intertemporal choices. This block also consists of a list of 126 fixed intertemporal choices.
• Again, both alternatives, A and B, consist of work schedules that distribute positive numbers of real-effort tasks over all 8 work periods. Again, upon completion of the respective work schedule, a payment of 100 euros is transferred to subjects’ bank accounts and subjects receive restaurant voucher.
• The alternatives depend on subjects’ Part-1 indifference points and on their (between-subjects) treatment condition. Details of the conditions are described below.
Part 3
In Part 3, subjects submit demographic information, complete a Raven IQ test, complete simple math tasks and complete the three classic CRT (cognitive reflection test) questions.

Between-subjects treatments:

Treatment Main

In each decision block of Part 1, there is exactly one block-defining work period for which subjects have to trade off an increase in work load with an increase in the voucher value. In treatment Main, subjects trade off an increase in the work load in all work periods for an increase in the voucher value.
More precisely, Alternative B is still constant, and its work schedule is the same as that of the Alternatives B faced in Part 1. Simultaneously, Alternative B in Part 2 goes along with a voucher value that is identical to the minimal voucher value included in all Alternatives B in Part 1. Alternative A, by contrast, goes along with a voucher value that is identical to the maximal voucher value across all Alternatives A in Part 1. Crucially, the work plans of the Alternatives A are created such that they include the work plan that is constructed from all eight indifference points elicited in Part 1. Let us refer to this particular work plan as the “reference work plan.” In all other choices, Alternative A includes a strictly greater or smaller number of tasks in each period.
This means that standard discounted utility would predict for the treatment Main in Part 2 that subjects are exactly indifferent between Alternative A for the “reference work plan” and Alternative B. The focusing model by Kőszegi & Szeidl QJE 2013), by contrast, predicts that subjects choose Alternative A even for amounts of work that are greater in every period than the indifference points elicited in Part 1 if focusing is strong enough.

Treatment Control

In treatment Control, one randomly selected choice from Part 1 is repeated. This allows us to detect—and control for—a potential time trend in subjects’ choices, which may be brought about by the fact that participants make similar decisions multiple times.

Treatment Main Dispersed

Treatment Main Dispersed operates just like Treatment Main with the exception of how the differences in the numbers of real-effort tasks per alternative and the differences in voucher values per alternative are displayed in Part 1 and Part 2 of the experiment.
In Treatment Main, the absolute number of real-effort tasks and voucher values for each alternative are displayed. Hence, the difference between real-effort tasks and voucher values have to be computed by the subjects.
In Treatment Main Dispersed, the differences in real-effort tasks are displayed by directly stating the difference of real-effort tasks between the alternatives. The differences in voucher values in Part 1 are also displayed that way. The differences in voucher values in Part 2 are displayed by splitting up the difference in 8 parts.

Treatment Control Dispersed

Treatment Control Dispersed operates just like Treatment Control with the exception of how the differences in the numbers of real-effort tasks per alternative and the differences in voucher values per alternative are displayed in Part 1 and Part 2 of the experiment.
In Treatment Control, the absolute number of real-effort tasks and voucher values for each alternative are displayed. Hence, the differences between real-effort tasks and voucher values have to be computed by the subjects.
In Treatment Control Dispersed, the differences in real-effort tasks are displayed by directly stating the difference of real-effort tasks and voucher values between the alternatives.

Treatment pairs (Main and Control or Main Dispersed and Control Dispersed) are randomized within sessions of up to 32 subjects on the individual level. We stratify treatment assignment among subjects within a session that do not meet the exclusion criterion. We do so by computing subjects’ average indifference point of Part 1, building an average indifference point ranking and randomizing pairs of direct-ranking neighbors into Main or Control (Main Dispersed or Control Dispersed). In alternating order, a session features either the treatment pair Main and Control or the treatment pair Main Dispersed and Control Dispersed.

Individual level

No

100 subjects in the treatment Main yield one indifference point each;
100 subjects in the treatment Control yield one indifference point each;
100 subjects in the treatment Main Dispersed yield one indifference point each;
100 subjects in the treatment Control Dispersed yield one indifference point each

100 subjects in the treatment Main yield one indifference point each; 100 subjects in the treatment Control yield one indifference point each; 100 subjects in the treatment Main Dispersed yield one indifference point each; 100 subjects in the treatment Control Dispersed yield one indifference point each

100 subjects in the treatment Main yield one indifference point each;
100 subjects in the treatment Control yield one indifference point each;
100 subjects in the treatment Main Dispersed yield one indifference point each;
100 subjects in the treatment Control Dispersed yield one indifference point each