Common Ownership and Firm Strategic Behaviour – Endogenous Ownership and Incentives

In our laboratory experiment two subjects in the role of managers will decide how much to produce in a Cournot market. The companies which they manage are owned by two subjects in the role of shareholders. Managers can choose to either produce the quantity that is the optimal choice under Cournot competition or half of the optimal monopoly quantity. The firms sell differentiated goods and a change in the own firm’s production has a larger effect on a firm’s price than a change in the competitors production.
We run two treatments that differ in the way that shareholders can incentivise managers.
In the first treatment, shareholders can choose to either:
1. Pay a fixed bonus regardless of their firm’s performance
2. Pay a bonus only if the profit of their firm is at least as high as that of the competing firm
3. Pay a bonus only if the margin of their firm is at least as high as that of the competing firm
The first choice has no effect on managers incentives. The second makes it a (weakly) dominant strategy to choose the Cournot quantity and the third makes it a (weakly) dominant strategy to choose half of the Monopoly quantity.
Incentives are chosen and revealed to the manager before the manager decides which quantity to produce. The manager’s salary is not paid out of the firms profit but is covered by the experimenter. Hence, a higher payment to the manager does not reduce the payment to the shareholders.
Our goal with this intervention is to see if shareholders choose incentives that are in their best interest and if managers then act in line with their incentives.
In our second treatment, shareholders cannot set ex ante incentives but can instead pay a voluntary bonus after observing the managers choice which may influence the managers decision in the next round. Our goal with this treatment is to see if a less formalised incentivisation mechanism that relies on repeated interaction has similar effects as formalised incentives. This is interesting because they are informal ways beyond ex ante agreed on incentive contracts through which managers and shareholders can interact in real world scenarios. For instance, shareholders can increase or decrease the managements base salary for follow-up contracts which can be seen as an ex-post bonus for good performance.
In both treatments we also study whether shareholders choose to exchange shares. The experiment is set up in a way that it is in the shareholders best interest hold shares in both firms and then incentivise the manager to produce halve the monopoly quantity. In both treatments the default is that shareholders start out by owning only shares in one firm. In future treatments we may also test different defaults such as shareholders starting out diversified with the opportunity to exchange shares so that they each only own shares in one firm to see if and how much the default influences the diversification outcomes.

2022-01-13

2022-12-31

We have three primary outcomes, the production choice of the manager, the incentive choice (either ex ante contract or ex post bonus) of the shareholder and the decision to exchange shares of the shareholder.

We directly record the produced quantity for each manager. They can make a binary choice between producing half of the monopoly quantity and the Cournot quantity.
We also record which incentive contract shareholders choose in treatment 1 and their bonus payment in treatment 2.
Finally, we record for each shareholder if they wanted to exchange shares in each round.
We are not only interested in the frequency with which each outcome is chosen but also with the relation between the outcomes. Therefore, we will study:
• The relationship between a manager’s production in a round and the bonus she receives in this round, depending on whether the shareholders are diversified or not
• The relation between the probability that the manager produces a different quantity than in the last round depending on the bonus she received in this round
• The production choice of the manager, depending on the incentive contract the shareholder picked
• The incentive contact the shareholder chooses depending on whether an exchange of shares happened or not

The experiment will be computerised, and subjects will be recruited from the online subject pool of the CESS lab at Oxford. There are two treatment groups which are described in detail in the previous sections. There will be 50 groups per treatment, each containing 2 shareholders and 2. Hence, we will recruit 200 subjects per treatment and 400 in total. Based on the experience with the subject pool we expect some subjects to not show up to the sessions. We compensate for this by inviting slightly more subjects. Because of this the total number of subjects may be slightly below or above the target number.

We will recruit different sessions and each session will contain subjects in both treatments. Subjects will be assigned to treatment groups based on the time they finish the consent stage of the experiment. Groups will be assigned to treatments 1 and 2 in alternating order. All Sessions will be recruited from the same population.

Subjects are randomly assigned to treatments. All subjects stay in the same matching group of four (two managers, two shareholders) for the whole session, so each matching group yields one independent observation.

No

No Clusters

We run our experiment with 50 groups of 2 managers and 2 shareholders each per treatment or 100 groups in total. The experiment has 15 rounds. This means that for the production decisions we have 2*15*100= 3000 observations. For the decision to choose and incentive contract and the decision which bonus to pay we have 15*100 = 1500 observations as these are only observed in one of the treatments each. For the decision whether to exchange shares we have 3000 observations like for the production decision and we will also analyse this separately by treatment. Because we collect multiple observations per subject and subjects are paired together into groups not all observations are independent. We will account for this by clustering standard errors on the subject level. We will also run regression specifications with subject fixed effects.

We plan that each treatment will have a sample size of 50 groups including 100 managers and 100 shareholders.

We did not do a power calculation for this experiment