Do intraorganizational transfers reduce Year-End Wasteful spending in organizations?

All explications are in Hidden.

I have five conditions :
EXPIRING
ROLLOVER
TR-SYMMETRIC
TR-ASYMMETRIC
TR-STRANGER

The two first treatments are mere problem decisions, i.e. there
are not a game. It means that the decisions made by a participant has no influence on
the paired participant’s decisions or payoffs. These treatments are the EXPIRING and the
ROLLOVER condition. In EXPIRING, budget leftovers at the end of the second semester
expires and cannot be used anymore. In ROLLOVER, budget leftovers are fully reported to
the next year, in addition of the yearly budget. Everything else is identical across treatments.

The three transfer (TR-) treatments are games as they all allow for interaction.
The treatments are called TR-SYMMETRIC,TR-ASYMMETRIC and TR-STRANGER. In these treatments,
every semester begins with a potentially transfer to the paired participant. The
decision is carried simultaneously and immediately implemented, so that the changes in
assets are relevant for the investment decision in projects in the same semester. The transfer
decision is also visible in semesters’ feedback information and in the history table. Otherwise, the
TR-SYMMETRIC condition is equivalent to EXPIRING. The ASYMMETRIC treatment’s core is an imbalance in the probability of getting high-
return projects across the pair members although the budget remains the same for both,
i.e. B = 5. More precisely, the more productive participant faces 0 high-return projects with pr. 10%, 1 with pr. 30% and 2 with pr. 60% while the less productive participant faces 0 high-return projects with pr. 40%, 1 with pr. 20% and 2 with pr. 40%. The expected expenditure in high-return projects for the more productive type is 6 while the expected expenditure in high-return projects for the more productive type is 4. So that they expected expenditure in high-return projects across both remains 5, like in the other treatments. Any other factors remain the same. This parametrization leads to the prediction that the less productive type should NOT transfer money to the more productive type, even if there would be efficiency gains in the transfer because the less productive type would forgo too much utility by doing this as the high productive type has never the incentive to do so.

The STRANGER treatment consists in rending impossible a repeated interaction. Participants are paired for one year and are rematched after any of them. The rematching occurs in matching groups of six participants. The history shows only the decisions of matched participants during the match. Neither they know with whom they are matched and whether they interacted already interacted before. Otherwise, the condition share the same features as TRANSFER. By construction, there is no bilateral reciprocity possible in STRANGER so that any transfer is feasible only if the weight for efficiency concerns is sufficiently high.

Each subject will face only one condition, i.e. we have a between-subject design.

2025-11-04

2026-01-20

Amount of low-return projects pursued
Amount of high-return projects pursued
Binary decision to transfer or not
Discounted social welfare (or funder function)

The discounted social welfare will be constructed as follows for all treatments except ROLLOVER: it is the difference between the value of a high-return project and its costs times the number of high-return projects pursued. (It is thus a monotonic transformation of the number of high-return projects pursued).

For the treatment Rollover, there will be a penalty on stored funds from previous year. This penalty represents the opportunity costs of the money not being able to be used elsewhere.

The penalty is a percentage between 0 and 100%. In the experiment the following percentage will be used:

0% (no penalty),
20% (low penalty),
50% (medium penalty),
80% (high penalty).

Risk preferences (RP)
Social preferences (SP)
Standard demographic question including a self-assesment of the strategic uncertainty and preference elicitation.

RP elicitation using the BRET from Close to and Flipping (2013)
SP elicitation using distributional preferences from Balafoutas et al. (2012)

All explications are in Hidden.

I design an indefinitely repeated project investment problem. The experiment has two goals: First, testing rules’
efficiency in reducing wasteful spending and in allowing available high-return projects to be
pursued. Second, testing the mechanisms behind transfer decisions. Namely, whether repetition
and symmetry are necessary and sufficient conditions for transfers to be sustained. Each treatment
changes one dimension of the TRANSFER condition. In every treatment, each participant
is paired to another. They take their decision simulatneously and can see subsequently the
decisions of their partner as well as their respective numbers of available high-return projects.
At any time, they see a history of all precedent decisions pursued by them and the matched
person. At the beginning of the year, both units receive a fixed and same budget.

At the beginning of any semester, subjects first learn the number of high-return projects which are available to them as well
as to their matched partners. They also observe their current assets and their of their partners. Then, each participant enters the number of high-return and low-return projects they want to pursue in this semester. The availability of high-return projects is stochastic: 0 with pr. 25%, 1 with probability 25% and 2 with probability 50%. The number has to be an integer and fulfill the two constraints about the availability of high-return projects and the non-negativity of assets for any semester. At the end of the semester, feedback about own and paired participant’s decisions and profits.

I use the block random termination design created by Fréchette and Yuksel (2017) with a block length of fourty periods and a continuation probability of 39/40 (97.5%). If the problem does not end during the first block, whether the games terminated or not will be announced period by period, starting with the fourty-first period (as pionneried in Vespa and Wilson (2019)).

The annual budget of each participant is the same in all treatments. This budget corresponds to the expected expenditure in high-return projects, as it is in theinterest of the funder to give as much money as necessary to pursue high-return projects
but also as less money as possible to avoid wasteful spending or long storage time in ROLLOVER. Formally, the budget that is allowed to units is 5 tokens. a high-return project costs 2 and a low-return project 1.

The hypotheses are the following:
1) The amount of low-return projects pursued has the following ordering: EXPIRING >= TR-SYMMETRIC > ROLLOVER

Expiring should be wasteful by design as their is no other way to reallocate resources. Rollover should be the least wasteful because there is the possibility to reallocate tokens to oneself in the future and TR-SYMMETRIC should be (weakly) better than Expiring as when rationalizable transfers are possible, there should be transfers.

2) The amount of high-return projects pursued has the following ordering: ROLLOVER >= TR-SYMMETRIC >= EXPIRING

ROLLOVER should provide the most appealing regime to induce efficient projects. TR-SYMMETRIC could be as efficient if absolutely everybody is pursuing rationalizable transfers. On the other hand, if nobody sends any transfers, TR-SYMMETRIC is then not better (but also not worse) than EXPIRING. This is why the inequality above are not strict.

3) The decision of pursuing a transfer (intensive margin) has the following ordering: TR-SYMMETRIC > TR-ASYMMETRIC = TR-STRANGER.

in TR-SYMMETRIC, both individuals who are selfish or efficiency-minded would transfer, since transfers are rationalizable for them. In the two other treatments, only efficiency-minded people would do it.

4) For sufficiently high penalty on stored funds in ROLLOVER, the discounted social welfare would give the aggregate results: TR-SYMMETRIC > ROLLOVER

For relatively high penalty on stored funds, the funder would prefer a budget regimes allowing transfers than commiting to rollover because of the opportunity costs of stored money.

Randomization by a computer

The level of radomization is group of 6 at the laboratory. The reason is because the stranger treatment has a matching group of 6 people, while the interaction is always in pair.

Yes

260 clusters in total

60 by treatment where their are fixed pairs throughout the experiment (EXPIRING, ROLLOVER, TR-SYMMETRIC, TR-ASYMMETRIC)
20 clusters in TR-Stranger because of matching groups of 6 participants.

600 participants

120 subjects in EXPIRING
120 subjects in ROLLOVER
120 subjects in TR-SYMMETRIC
120 subjects in TR-ASYMMETRIC
120 subjects in TR-STRANGER

0.1 SD (calculated with ICC = 0.03). This holds for the four main outcomes. Including multiple hypotheses correction with romano-wolf stepdown procedure.