Grid-group cultural theory and behaviour in a public good experiment

ABSTRACT

This paper investigates the interaction between one’s cultural orientation, as derived from Grid-Group Cultural Theory (GGCT), and behaviour in a public good (PG) experiment with and without the opportunity to monitor and punish. We observe that high-group individuals contribute more than low-group individuals and that high-grid individuals are more likely to monitor and punish than low-grid individuals. Results confirm that one’s culture interacts in predictable ways with economic institutions. Future research is warranted to determine whether such institutions can be designed to appeal to specific cultures in order to achieve public policy goals more effectively.

KEYWORDS:

• Grid-group cultural theory
• public good experiment
• culture
• experimental economics
• cooperation

JEL CODES:

• C92
• H41
• D82

I. Introduction

Despite recognition that culture affects behaviour, the extent to which it interacts with economic incentives remains under-explored (Verweij et al. 2006). To investigate, we employ a public good (PG) experiment with and without the opportunity to monitor and punish and test hypotheses derived from Grid-Group Cultural Theory (GGCT). As articulated by Douglas (1982); Douglas and Wildavsky (1983), GGCT theorizes two dimensions, grid, ‘the degree to which our lives are circumscribed by conventions or rules’ and group, ‘the extent…individual choice is constrained by group choice, by binding the individual into a collective’ (Hood 1998, p. 8).¹

Public good experiments are often used to investigate cooperative behaviour since they create a stark social dilemma. Cooperative behaviour (contributing towards the public good) benefits the group but conflicts with one’s own self-interest. At the individual level, each group member has an incentive to avoid contributing (freeride) but at the group level it is optimal if each group member contributes fully. In the absence of institutional constraints on behaviour, what is known as the voluntary contributions mechanism (VCM), individual incentives dominate and contributions trend towards zero (Davis and Holt 1993; Ledyard 1995).

An alternative mechanism, peer punishment, allows participants to observe the contributions of other’s and to pay a fee to sanction them. While this mechanism may increase contributions, results are mixed and are sensitive to individual behaviour within groups (Fehr and Gächter 2000 & 2002; Chaudhuri (2011). Two contrasting instances stand out in the literature. First, when the use of punishment is insufficient to alter individual incentives, contributions remain suppressed and mirror what is observed in the VCM. This is not surprising since, as Yamagishi (1986) points out, the provision of costly peer punishment is, itself, a public good dilemma and, as such, conflicts with one’s self-interest. Second, when the use of punishment is excessive or perverse (the targeting of cooperative group members), it no longer encourages cooperative behaviour (Cinyabuguma, Page, and Putterman 2006; Ertan, Page, and Putterman 2009).

To better understand individual variation in response to economic and institutional incentives, we investigate the role of one’s cultural orientation. Chai et al. (2011) conducted a PG experiment with and without peer punishment to test whether behaviour correlated with postulates of GGCT. Using categories derived from the World Values Survey, they found high-group scores correlated with higher contributions (consistent with prioritizing group welfare) and high-grid scores correlated with increased willingness to punish (consistent with employing the institution to regulate the behaviour of others). Here, we use a post-experiment survey to measure one’s grid-group designation similar to Ostrander (1982), Hampton (1982), and Gross and Rayner (1985) to directly infer the effect of one’s cultural orientation on behaviour. Consistent with predictions, high-group individuals contribute more than low-group individuals and high-grid individuals are more likely than low-grid individuals to monitor and punish.

This research suggests one’s cultural orientation interacts, in predictable ways, with economic incentives within institutions designed to achieve policy goals. While we note the limitations of this research in the discussion, further research is warranted to determine whether institutions designed to appeal to specific cultural orientations more effectively achieve policy goals.

II. Method

This experiment implemented a PG experiment with groups of 4 across 33 periods and included 3 parts.² Part 1 (periods 1–9) introduces a voluntary contributions mechanism (VCM). Before each period, group members received a fixed endowment of experimental dollars (EDs). Each member then independently allocated their endowment across their private and a group account (the public good). Payoffs to subject i are:

${\pi }_i=\left(e_i-x_i\right)+\mathit{\alpha }\sum _{\mathit{j}=1}^n{x}_j$

Where:

x_i = subject’s contribution to the group account,

e_i = 20 is the subject’s endowment,

α = 0.5 is the marginal per capita return (MPCR) from the public good,

and $\sum _{\mathit{j}=1}^n{x}_j$ represents the sum of group account contributions.

With n players,$\frac{1}{n}$ < α < 1, and a known last period, the Nash equilibrium is that each member contributes nothing while the social optimum requires each member to contribute their entire endowment.

Part 2 (periods 10–27) introduces an opportunity to monitor and punish. After their allocation decisions, participants may choose to monitor individual group account allocations.³ Monitoring individual allocations is costly, the total cost of monitoring is 4 EDs. However, one’s individual cost depends on the number of group members willing to monitor such that the total monitoring cost is evenly split among all those choosing to monitor. Further, the opportunity to punish is not restricted to those who choose to monitor. Once monitoring is provided, all group members have the opportunity to punish. This design was chosen to differentiate between potential motivations within GGCT across the decision to monitor (providing the opportunity to punish) and the decision to punish if given the opportunity. Absent monitoring, individual payoffs are identical to part 1. Otherwise, payoffs to i are:

${\pi }_i=\left(e_i-x_i\right)+\mathit{\alpha }\sum _{\mathit{j}=1}^n{x}_j-m_i-\mathit{c}\sum _{\mathit{j}\ne \mathit{i}}^{\mathit{n}-1}{P}_{\mathit{ij}}-\mathit{r}\sum _{\mathit{j}\ne \mathit{i}}^{\mathit{n}-1}{P}_{\mathit{ji}}$

Where:

m_i = 4/n_m if i chooses to monitor and zero otherwise,

n_m = number of individuals choosing to monitor,

P_ij = number of reduction points i imposes on others j at a cost of c = 1 each, and

P_ji = number of reduction points other group members j impose upon i at a cost of r = 4 each.

Hypothesis 1

High-group individuals contribute more than low-group individuals, one’s grid orientation has no significant effect.

Hypothesis 2

High-grid individuals are more likely to monitor and punish than low-grid individuals, one’s group orientation has no significant effect.

III. Results

Fifty-two subjects (13 independent groups) participated in an hour-long experiment. Average earnings were $17.63, show-up payment included. To determine one’s cultural orientation, participants answered questions specifically designed to measure agreement with statements aligned to GGCT types in a post-experiment survey. Using these scores we characterized participants as low- or high-group and low- or high-grid. Participants scoring higher on high-group orientated, relative to low-group orientated questions were labelled high-group.⁴

IV. Contributions

The relationship between one’s grid-group designation and contributions is examined by interacting indicators for part, grid designation, and group designation (Table 1). To test our hypotheses, we determine average contributions for each grid-group designation in isolation and then estimate the marginal effect of the difference. For example, we first determine the predicted contributions from low and high-group individuals and then, holding grid designation constant, we estimate the predicted difference (Table 2). We observe one’s group designation has a significant effect on one’s contribution. In parts 1 and 2, high-group individuals contribute significantly more than low-group participants. Hypothesis 1 is confirmed and is consistent with high-group individuals’ prioritizing group welfare over self-interest.⁵

Table 1. Contributions.

	Contributions
Part 1/Low-Grid/High-Group	4.44**
	(1.98)
Part 1/High-Grid/Low-Group	1.32
	(2.17)
Part 1/High-Grid/High-Group	2.45
	(2.56)
Part 2/Low-Grid/Low-Group	2.26***
	(0.761)
Part 2/Low-Grid/High-Group	6.18***
	(2.16)
Part 2/High-Grid/Low-Group	−1.07
	(2.44)
Part 2/High-Grid/High-Group	6.03**
	(2.68)
Constant	8.65***
	(1.76)
N	1215
R^2	0.12

Individual random effects with errors clustered at group level.

*p < 0.10, **p < 0.05, ***p < 0.01.

Table 2. Grid-group effects.

	Low-Grid	High-Grid	Diff.	Low-Group	High-Group	Diff.
Cont. Part 1	11.91***	10.80***	−1.11	9.03***	12.52***	3.48**
	(0.842)	(1.60)	(1.73)	(1.31)	(0.935)	(1.56)
Cont. Part 2	13.79***	12.79***	−1.00	9.95***	14.79***	4.83***
	(1.28)	(1.91)	(1.82)	(1.38)	(1.45)	(1.64)
Prob. Monitor	0.141***	0.271***	0.129***	0.155***	0.184***	0.029
	(0.039)	(0.028)	(0.045)	(0.039)	(0.032)	(0.035)
Prob. Punish	0.541***	0.758***	0.217***	0.698***	0.579***	−0.119
	(0.063)	(0.049)	(0.064)	(0.066)	(0.067)	(0.098)
Punishment	1.87***	3.17***	1.30**	2.76***	2.11***	−0.648
	(0.249)	(0.560)	(0.621)	(0.586)	(0.222)	(0.592)

Figures represent predicted averages within designation, standard errors in parenthesis.

*p < 0.10, **p < 0.05, ***p < 0.01.

IV.I. Monitoring and punishment

As above, we investigate the relationship between using the monitoring and punishment mechanism and grid-group designation using interaction terms; we also include period to account for time trends, one’s own contribution, and the sum of contributions to the PG (Table 3).⁶

Table 3. Monitoring and punishment.

	Prob. Monitor	Prob. Punish	Punishment
Period	−0.073***	−0.014	0.028
	(0.015)	(0.015)	(0.039)
Contribution	0.063*	0.067***	0.134***
	(0.033)	(0.016)	(0.029)
Sum of Contributions	−0.011	−0.024***	−0.056***
	(0.01)	(0.009)	(0.014)
Low-Grid/High-Group	0.369*	−0.369	0.024
	(0.215)	(0.370)	(0.659)
High-Grid/Low-Group	1.02***	0.908	2.89*
	(0.302)	(0.665)	(1.648)
High-Grid/High-Group	0.845***	0.349	0.781
	(0.298)	(0.418)	(0.571)
Constant	−0.564	1.05**	2.65***
	(0.44)	(0.44)	(0.581)
N	810	406	406
X^2	34.86	106.04	35.70

Individual random effects with errors clustered at group level.

*p < 0.10, **p < 0.05, ***p < 0.01.

The probability of monitoring decreases across periods. We observe a positive relationship between one’s contribution and monitoring probability, suggesting those making higher contributions are more likely to monitor and provide the opportunity to punish. The probability of punishing and amount of punishment increase with one’s contribution and decrease as the sum of contributions increases.

Next, we combine interaction terms to determine how one’s grid-group designation impacts monitoring and punishment behaviour (Table 2). We observe that one’s group designation does not impact their use of monitoring or punishment, while one’s grid has a significant effect. High-grid (relative to low-grid) individuals are more likely to monitor (27% vs. 14%) and to punish (76% vs. 54%). High-grid individuals also impose more punishment than low-grid individuals (3.17 vs. 1.87). This confirms our second hypothesis and is consistent with a willingness to employ institutional features to regulate others’ behaviour.

V. Discussion

This research investigates the impact of one’s cultural orientation on behaviour within a PG experiment with and without the opportunity to monitor and punish. In parts 1 and 2, high-group individuals contributed significantly more than low-group individuals. In part 2, high-grid individuals were more likely to monitor, punish, and punish more than low-grid individuals. Results support our hypotheses and confirm the correlations observed in Chai et al. (2011), suggesting that individual behaviour within institutions is influenced by one’s cultural orientation.

It is important to note the limitations of using an ex-post elicitation of cultural orientation. However, these results are encouraging and further research is warranted to determine whether institutions designed to appeal to specific cultural orientations more effectively achieve public policy goals. Of particular importance would be to elicit cultural orientation prior to participation in the experiment. This would allow researchers to sort participants into culturally specific groups to determine the impact of culture at the group level.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the University of Massachusetts Lowell.

Notes

¹ GGCT therefore posits 4 cultures: hierarchist (high-group/high-grid), fatalist (low-group/high-grid), egalitarian (high-group/low-grid), and individualist (low-group/low-grid).

² Here, we discuss only parts 1 and 2.

³ Instructions for part 2 (including comprehension questions) were provided after the completion of part 1.

⁴ Seven participants were removed from the regression analysis because we were unable to discern their GGCT (e.g. answering each question with ‘4’, on a 7 point Likert scale, prevents discerning a participant’s designation).

⁵ Disregarding grid-group designations, average contributions across part 1 (11.63*** (0.776)) and part 2 (13.46*** (1.23)) are significantly different (1.83** (0.936)). Consistent with the literature, the opportunity for punishment increased group account contributions.

⁶ The discrete decisions to monitor and punish are analysed with a probit specification.

Appendices

Cultural Theory Battery

I would like to understand more about how you feel about American society. On a scale from one to seven, where one means you strongly agree, and seven means you strongly disagree, please respond to each of the following statements.

Egalitarian (High-Group/Low-Grid)

• Society works best if power is shared equally.

• It is our responsibility to reduce the differences in income between the rich and the poor.

• What our society needs is a fairness revolution to make the distribution of goods more equal.

Fatalist (Low-Group/High-Grid)

• It would be pointless to make serious plans in such an uncertain world.

• No matter how hard we try, the course of our lives is largely determined by forces outside our control.

• Most of the important things that take place in life happen by random chance.

Individualist (Low-Group/Low-Grid)

• We are all better off when we compete as individuals.

• Even the disadvantaged should have to make their own way in the world.

• Even if some people are at a disadvantage, it is best for society to let people succeed or fail on their own.

Hierarchist (High-Group/High-Grid)

• Society would be much better off if we imposed strict and swift punishment on those that break the rules.

• Our society is in trouble, because we don’t obey those in authority.

• The best way to get ahead in life is to do what you are told to do.

Welcome to the Experiment

Thank you for participating in our decision making experiment. The experiment has 2 parts. Part 1 consists of 10 periods and is explained below. We will explain part 2 following the completion of part 1. In each period you will have an opportunity to earn money, which is in addition to the $5 guaranteed for your participation.

Please read the following instructions carefully. Everyone must correctly answer the comprehension questions on page 3 before we can begin.

During the experiment you are not allowed to communicate with other participants. If you have a question, please raise your hand.

During the experiment your earnings will be calculated in Experimental Dollars (EDs). You can earn EDs every period. Your total earnings are the summation of your period earnings. At the end of the experiment, your total earnings in EDs will be converted to U.S. dollars at the following rate:

EDs = $1

At the end of the experiment your total earnings (including the $5 participation payment) will be paid to you, privately and anonymously, in cash.

In the experiment, each participant is randomly assigned to a group of 4. You will be part of the same group throughout the entire experiment. However, at no point will the members of your group be revealed. All of the decisions you make within the experiment are anonymous – no identifying information will be collected.

In every period, each group member, yourself included, will be given an endowment of 20 EDs.

Summary

You are in a group of 4.

Each group member begins each period with 20 EDs.

Your total earnings equal the summation of your period earnings.

Part 1

Each period in part 1 consists of a single stage (stage 1). In stage 1, each of you will decide how many of your EDs to allocate to a group account. You can allocate any integer between 0 and 20 to the group account. Your remaining EDs will be kept in your private account.

How are stage 1 earnings calculated?

Your stage 1 earnings depend on the number of EDs in your private account and the total number of EDs in the group account.

Your Stage 1 Earnings =

(20 – Your Allocation to the Group Acct.) + 0.5*(Total Number of EDs in the Group Acct.)

In stage 1 you will be shown your group account allocation, the total group account allocation, and your stage 1 earnings.

Examples

1. Assume the following allocations to the group account from each group member.

Table

Group Member	Endow.	Alloc.	Private Acct. Earnings	Total EDs in Group Acct.	Group Acct. Earnings	Stage 1 Earnings
			1. (20–Alloc.)	2. (10+5+15+20=50)	3. (0.5*Total)	4. ((1.)+(3.))
You	20	10	10	50	25	35
Member A	20	5	15	50	25	40
Member B	20	15	5	50	25	30
Member C	20	20	0	50	25	25

How solve?

1. Find each member’s earnings from their private account (20 – their allocation)

2. Find the total number of EDs allocated to the group account (10 + 5 + 15 + 20 = 50)

3. Find each member’s earnings from the group account (0.5 × 50 = 25)

4. Find each member’s stage 1 earnings (private acct. earnings + group acct. earnings)

Note that for any given amount of EDs in the group account, the lower one’s allocation to the group account the higher their stage 1 earnings. For example, member C allocated 20 and earned 25 EDs while member A allocated 5 and earned 40 EDs.

This is because each group member earns the same amount from the group account (here 25 EDs) but keeps all the EDs left in their private account.

2. Assume the same allocations to the group account as above except that your allocation decreases from 10 to 0.

Table

Group Member	Endow.	Alloc.	Private Acct. Earnings	Total EDs in Group Acct.	Group Acct. Earnings	Stage 1 Earnings
You	20	0	20	40	20	40
Member A	20	5	15	40	20	35
Member B	20	15	5	40	20	25
Member C	20	20	0	40	20	20

What has changed from example 1?

1. Your private account earnings increased from 10 to 20

2. The total EDs in the group account decreased from 50 to 40 (0 + 5 + 15 + 20)

3. The group account earnings of each member decreased from 25 to 20

4. Your stage 1 earnings increased by 5

5. The stage 1 earnings of each other member decreased by 5

Stage 1: Comprehension Questions

Please answer the following questions. Raise your hand if you have a question.

1. Assume the following allocations to the group account from each member. Fill in table.

Table

Group Member	Endow.	Alloc.	Private Acct. Earnings	Total EDs in Group Acct.	Group Acct. Earnings	Stage 1 Earnings
You	20	20
Member A	20	5
Member B	20	15
Member C	20	20

1. What are the stage 1 earnings of each member if each member allocates 0 EDs to the group account? Use the formula on Page 2.

2. What are the stage 1 earnings of each member if each member allocates 20 EDs to the group account? Use the formula on Page 2.

Part 2

Part 2 consists of 18 periods. Each period in part 2 has two stages. Stage 1 is exactly as it was in part 1. You each will remain in the same group and make the same allocation decision. However, stage 2 of each period is new.

Stage 2

After stage 1 you will continue to be shown your group account allocation, the total group account allocation, and your stage 1 earnings.

Immediately following stage 1 each group member will be asked (Yes or No) whether they would like to observe the individual group account allocations of each group member. Observing individual group account allocations is costly.

If at least one member decides to observe the individual allocations the experiment will continue with stage 2 (explained below). Otherwise, if no member decides to observe the individual allocations stage 2 will conclude and the experiment will continue to the next period.

How are individual monitoring costs calculated?

The total cost of observing individual allocations is 4 EDs. However, the individual monitoring cost is the total cost divided by the number of members who chose to observe individual allocations as follows:

Table

Number of group members who decide to observe individual allocations	Total Monitoring Costs (EDs)	Individual Monitoring Costs (EDs)
1	4	4.00
2	4	2.00
3	4	1.33
4	4	1.00

If you chose not to observe individual allocations than your individual monitoring costs will always be 0 EDs.

If individual allocations are observed each group member will be shown the group account allocations of each group member by random ID. This means that the order of presentation of individual allocations is randomly shuffled each period.

In stage 2, you can decide whether or not to reduce the earnings of other group members. Each group member will have the opportunity to assign Reduction Points (RPs). You can assign up to 3 RPs to each group member. Reductions points are costly to impose and costly to receive.

For each RP you impose on another member you will pay 1 ED.

For each RP you receive from another member you will pay 4 EDs.

How to determine the costs of imposing RPs for each member?

Example

Assume that after observing individual allocations, the following RPs are imposed.

Table

Group Member	Reduction Points Imposed	Reduction Points Received
You	1 RP on Member C	0 RPs
Member A	2 RPs on Member C	1 RP from Member B 1 RP from Member C
Member B	2 RPs on Member C 1 RP on Member A	1 RP from Member C
Member C	1 RP on Member A 1 RP on Member B	1 RP from you 2 RPs from Member A 2 RPs from Member B

Each RP imposed costs 1 ED.

1. What are the costs of imposing RPs for you?

You imposed 1 RP on member C, your cost of imposing 1 RP = 1 ED.

2. What are the costs of imposing RPs for member A?

Member A imposed 2 RPs on member C, the cost of imposing 2 RPs = 2 EDs.

How to determine the costs of receiving RPs for each member?

Each RP received costs 4 EDs.

1. What are the costs of receiving RPs for you?

You received 0 RPs, your cost of receiving 0 RPs = 0 EDs.

2. What are the costs of receiving RPs for member A?

Member A received 1 RP from member B and 1 RP from member C.

The cost of receiving 2 RPs = 8 EDs.

Stage 2 Comprehension Questions

Please answer the following questions using the example. Raise your hand if you have a question.

1. What are the costs of imposing RPs for member B?

2. What are the costs of imposing RPs for member C?

3. What are the costs of receiving RPs for group member B?

4. What are the costs of receiving RPs for group member C?