Intervention (Hidden)
Our experimental design is based on trust, which is a vital ingredient for economic exchange and development. We use a trust game, which is the cornerstone of cooperation dilemmas and for measuring trust as a cultural trait. The game has two players, a trustor and a trustee. We give both players the same endowment at the beginning of the game. Crucially, the players take decisions sequentially: the trustor decides first to send any amount between zero and her endowment to the trustee, which is tripled by the experimenter. The trustee then decides whether and how much to transfer back to the trustor. The total payoffs are maximized if the trustor sends her entire endowment in the expectation that the trustee reciprocates. However, if the trustee does not, the trustor loses money. This creates a cooperation dilemma.
The trust game has three implications that are relevant for isolating the role of anticipation of discrimination from confounding motives: a) Since the trustor moves first, her decision may reflect anticipation of discrimination, as well as taste-based and statistical discrimination. b) Since the trustee is the second mover, she can condition her decision on the trustor's transfer. This implies that the trustee's decision is driven by taste-based discrimination, but not statistical discrimination or anticipation of discrimination. c) Since the trustee can condition her decision on the trustor's transfer and the trustor knows that, the game rules out higher-order beliefs (anticipation of anticipation, and so on) from playing a role, which simplifies the challenge of separating anticipation of discrimination from other motives of the trustor. In addition, the behavior of trustees can be used to shed light on whether anticipation of discrimination by trustors is misperceived.
To separate the different motives underlying a trustor's behavior, we augment the trust game as follows. We randomly match individuals with an ingroup or outgroup person, so that individuals are aware that there is a 50 \% chance of being matched with an outgroup person. Then we experimentally vary the information structure of the game. We flip a coin for each individual within a match to determine whether her own identity is revealed or concealed to the other person, such that the outcomes of the coin flip are common knowledge. This generates four key treatments: outgroup-revealed, ingroup-revealed, outgroup-concealed, and ingroup-concealed.
We take the \textit{first} difference ($D_{1}$) between the amount sent by the trustors to outgroup vs. ingroup trustees, when the trustor's own identity is \textit{revealed}. The first difference captures the importance of taste-based and statistical discrimination by the trustor, as well as her anticipation of discrimination, as the trustor's own identity is revealed. We then take the \textit{second} difference ($D_{2}$) between the amount sent by the trustors to outgroup vs. ingroup trustees, when the trustor's own identity is \textit{concealed}. The second difference reflects taste-based and statistical discrimination by the trustor but cannot be influenced by anticipation of discrimination, as the trustor's own identity is concealed. Since taste-based and statistical discrimination by the trustor should not depend on whether her own identity is revealed or concealed, a double difference between $D_{1}$ and $D_{2}$ allows us to estimate the importance of anticipation of discrimination over and above a trustor's taste-based and statistical discrimination. We also elicit a trustor's expectation of the amount they think the trustee will return. This allows us to go beyond the reduced-form evidence and confirm whether anticipation of discrimination indeed undermines trust.