Promoting Intercaste Harmony Through Sports

1,261 men in rural India
2017 - 2019
Target group:
  • Men and boys
  • Rural population
  • Adults
Outcome of interest:
  • Discrimination
  • Attitudes and norms
  • Social cohesion
Intervention type:
  • Community participation
  • Social networks
  • Intergroup/social contact
AEA RCT registration number:
  • George and Obie Shultz Fund

Contact between members of different social groups might have the potential to reduce prejudice and promote intergroup harmony. However, little rigorous evidence exists on the conditions under which this might be the case. Researchers conducted a randomized evaluation of cricket teams in rural India to study the impact of members of different castes either playing on the same team (collaborative contact), or in rival teams (adversarial contact), on cross-caste prejudice and interaction. Collaborative contact between cricket players of different castes reduced prejudice toward members of lower castes and increased cross-caste interaction, while adversarial contact increased prejudice in intergroup interactions.

Policy issue

Contact between different social groups may have the potential to reduce prejudice. However, little rigorous evidence exists on the conditions under which this might be the case. For example, while teamwork among members of different social groups may promote social cohesion (collaborative contact), competition between groups with opposing goals could exacerbate their rivalry and prejudice (adversarial contact). By understanding which types of intergroup contact have positive or negative effects, policymakers may be able to design optimal policies for integration, and combat prejudice and discrimination toward disadvantaged groups.

Context of the evaluation

In India, prejudice and discrimination toward lower castes are widespread. India’s caste system divides individuals into four ordered social categories: Brahmins, Kshatriyas, Vaishyas, and Shudras, with the lowest group, the “untouchables”, outside of the hierarchy. While norms of contact between the castes largely dictate social relations, such as segregated residence areas and social networks, individuals are also divided into three government categories: General, Other Backwards Castes (OBC), and Scheduled Castes / Scheduled Tribes (SC/ST). 

Although discrimination toward lower castes is illegal, 39 percent of General and OBC households in the study setting of Uttar Pradesh still practice “untouchability” (compared to 24 percent in India) by reducing physical interactions with lower castes to remain “unpolluted.” In the setting of this evaluation, participants were nearly twice as likely to form friendships with a participant of the same caste than with a member of any caste in general. Similarly, 50 percent of Uttar Pradesh households reported substantial conflict between different caste groups in their village.  

To examine types of social contact in this setting, researchers used cricket, a team-based, bat-and-ball sport that is widely popular across social groups in India. The sport is a common and natural source of social contact in Uttar Pradesh: 81 percent of study participants played cricket at least two times per week, and 38 percent were aware of a local cricket tournament in the past year.  

Men in India play cricket outdoors
Photo: Salvacampillo |

Details of the intervention

Researchers conducted a randomized evaluation of cricket leagues in the state of Uttar Pradesh, India, to test the effects of two types of intergroup contact – collaborative and adversarial – on cross-caste prejudice and interaction. Researchers recruited 1,261 men, aged 14 to 30, who were interested in participating in cricket leagues across eight gram panchayats (GPs, or local administrative units) in India. They randomly assigned the men to either play in the cricket leagues and encounter two types of social contact, or to a comparison group: 

  • Collaborative contact: Out of the 1,261 men, 800 were randomly assigned to play in eight cricket leagues, one in each GP. Each cricket league lasted for a month, with matches occurring from January to July 2017. In every cricket league, 35 percent of players (7 out of 20 teams) were randomly assigned to homogenous-caste teams (with all players from the same caste), and the others to mixed-caste teams (with players from different castes). This strategy allowed researchers to assess whether being on the same team as other-caste participants (collaborative contact) affected the players’ prejudice and cross-caste interactions.

  • Adversarial contact: In each cricket league, all homogeneous- and mixed-caste teams played eight matches against randomly selected opponents. This allowed researchers to assess the impact of playing against members of a different caste (adversarial contact) on prejudice and interactions.

  • Comparison group: The remaining 461 players were selected to be back-up players for teams. Some were selected as “high-priority backups,” who came in to replace team members from their same caste and played most frequently. This left a group of “low-priority” backups who only played 1.6 matches each on average (compared with 6.1 for league players). Researchers used the “low-priority” backups as a comparison group, since they had much lower exposure to any type of social contact in the cricket leagues.

After recruitment was finalized and before the leagues began, researchers gathered data about players’ cricket abilities and their impressions of and friendships with other participants. After players were assigned to teams, they were also given their teammates’ full names and family names (a strong signal of caste) over the phone to allow them to implicitly identify their teammates’ castes. 

Surveyors recorded players’ interactions at each match. One to three weeks after each league ended, researchers used surveys and additional exercises to examine whether the intervention affected participants’ prejudice and interactions along three main areas. First, researchers asked participants about their willingness to interact with players of different castes, as well as their cross-caste friendships. Second, researchers measured participants’ favoritism for their own caste by asking participants to vote anonymously for one member of each team to receive professional cricket coaching. As participants were told to vote for the player they most preferred, this voting exercise allowed researchers to estimate the effect of contact on caste favoritism (e.g. players voting for others based on their caste rather than their ability).

Finally, in a separate exercise, researchers measured whether the intervention had an effect on cross-caste trading. After the leagues ended, researchers gave all participants two goods – a pair of gloves and a pair of flip-flops – that were intentionally mismatched. As all received either two left-hand or two-right hand gloves, and two left-foot or two right-foot flip-flops, they were offered a monetary incentive of INR 10 to 20 (US$0.16 to $0.32) to make a successful trade with another participant to correct the mismatch. Researchers also placed colored stickers on each mismatched pair that correlated strongly with their recipient’s caste, and offered a randomly-selected half of participants extra monetary incentives of INR 50 to 100 (US$0.78 to $1.56) for “color-switching” – or trading for an object with a different color sticker. This allowed researchers to test whether collaborative vs. adversarial contact from the matches affected the likelihood of players trading with each other, with and without the presence of stronger monetary incentives for cross-caste trade.   

Results and policy lessons

Researchers found that collaborative contact with teammates from different castes reduced prejudice and increased cross-caste interactions, whereas adversarial contact with other-caste opponents tended to have the opposite effect. Overall, league participation reduced prejudice relative to the comparison group, which suggests that the positive impacts of intergroup contact outweighed the negative effects of adversarial contact. 

Willingness to interact and friendships: Collaborative contact increased desired interactions and friendships with members of different castes, while adversarial contact had the opposite effect. Having all other-caste teammates instead of none increased desired number of interactions with players from other castes by 2.2 (a 31 percent increase, compared to an average of 7.1 desired interactions among homogenous-caste teams) and increased cross-caste friendships by 1 (a 32 percent increase, compared to an average of 3.1 cross-caste friendships amongst homogenous-caste members). As collaborative contact increased friendships even when players had low cricket abilities (or won fewer games), this suggests that the positive effects of collaborative contact come from working, not winning, together. In contrast, moving from the lowest (35 percent) to highest proportion (97.5 percent) of other-caste opponents decreased desired interactions with players from other castes by 4.9 (a 69 percent decrease, compared to an average of 7.1 desired interactions among homogenous-caste team members), and reduced cross-caste friendships by 2.2 (a 71 percent decrease, compared to an average of 3.1 cross-caste friendships among homogenous-caste members). Relative to the comparison group, league participants (either in the collaborative or adversarial group) increased their cross-caste friendships by 1.1 by the end of the period (a 44 percent decrease, compared to an average of 2.5 friendships among non-players) suggesting the positive impacts of collaborative contact outweighed the negative impacts of adversarial contact.

Caste Favoritism: Collaborative contact decreased favoritism in the voting exercise, while adversarial contact did not have a significant effect. On average, players ranked participants of their same caste 0.4 positions higher along a scale of 1 (bottom) to 5 (top) that determined who won the prize. Collaborative contact decreased the amount of favoritism by reducing the 0.4 advantage by a third (or 0.13).  

Trade: League participation (under any type of contact) increased the quantity of cross-caste trade by 8.4 percentage points, relative to the comparison group average of 49.1 percent of items traded. Overall, most non-backup league players traded their items successfully, with no significant difference between participants who experienced collaborative contact or adversarial contact. However, for the half of players who were offered additional “color-switching” incentives for cross-caste trade, collaborative contact increased cross-caste trade by 11 percentage points (a 21 percent increase, over an average of 52 percent of items traded by homogenous-caste teams)—and, consequently, trade payouts by 18 percent (an INR 15 (US$0.23) increase from an average payout of INR 83 (US$1.29)). Researchers estimated that this effect was the same as the predicted effect of offering participants one to two hours’ worth of local daily wages for trading with other-caste members. This suggests that collaborative contact can substantially facilitate cross-caste economic interaction in the presence of additional incentives. 

These results suggest that short-term sports programs can be effective in strengthening cooperation and reducing prejudice. Further, intergroup contact interventions may be most effective when integration takes place across small groups with common goals – such as cricket players working together on the same team.