Experimental Design Details
The incentivized online experiment with families from Saudi Arabia and Hong Kong aims to understand the behavioral change in individuals in response to the treatment in the RCT, where the treatment bundles together several potential mechanisms. We identify three channels through which the treatment effect might occur:
1) feeling of reciprocal obligations by the employers,
2) decreased social distance between employers and domestic workers (DWs),
3) and employers’ increased moral cost of causing harm to DWs.
These channels should result either in more altruistic feeling of employers towards the DWs and/or lead to reduced antisocial behavior. We also hypothesize that kind actions at the first instance of an interaction promote kind reciprocal behavior from the other party, relative to when the kind action only comes at a later stage.
We cannot merely learn from existing laboratory experiments, since while earlier literature has established that reciprocity and reduced social distance both enhance immediate prosociality towards individuals, it does not examine persistence of these effects and spillovers on other members of the same group (only Gneezy and List 2006 or Kube et al. 2012 document some short term persistence of response to gifts). Also, extrapolating and generalizing results from laboratory settings with undergraduate students can be misleading in the particular setting we study, where social norms guiding interactions of employers and DWs have been long established and hence guide individual behavior.
We will use incentivized experiments typically used in a laboratory that closely resemble the situation of Filipino domestic workers in Saudi Arabia and in Hong Kong: 1) the bargaining power is heavily skewed towards the employers, 2) the interaction is of fairly long duration but one-shot (one employment contract, assuming away possibility of reemployment as getting a new domestic worker is not more costly than prolonging a contract with a new one), and 3) there is a possibility for the domestic worker to send a signal to the employer at the first instance of their interaction: either through giving him or her a small gift (with no substantial monetary value but with a positive intrinsic value), or through increasing familiarity and reducing social distance by presenting a picture of their family. The signal is expected to induce employers to behave more prosocially towards the workers and employers’ moral cost of harmful action towards the domestic workers should increase.
We will use two games: a dictator game with a restricted budget set and an adapted version of two modified dictator games used in Fehr et al. (2008) and Bauer et al. (2014). Unlike in their case, we will not use two binary games, but rather we make the choice space “more continuous” by having multiple choices per category. In each game, a sender has to select between two alternative allocations of tokens for him/herself and the receiver.
The dictator game allows the sender to redistribute US$20 between him/herself and the receiver. The sender can then allocate anything between US$0 and US$10 units in increments of US$1 to the receiver using a slider bar.
Manipulating default endowment allocation (orthogonal to treatment)
Further, we will randomly manipulate whether the initial allocation on the slider bar starts at (20,0)—i.e. all for the sender and nothing for the sender—or whether it starts at an egalitarian split (10,10), allowing for “taking away from the receiver”. Although the choice space is theoretically equivalent, the framing results in different norms associated with the redistributive choices (Krupka and Weber, 2013; Cox et al., 2016). While a distribution (15,5) is perceived as rather generous gift of US$5 in the former game, taking US$5 away from the receiver in the latter game, is perceived as rather nasty. As such, we will manipulate the moral cost of harmful behavior independent of the role of social distance and reciprocity.
Measure: inequality aversion (Bolton and Ockenfels 2000; Fehr and Schmidt 1999) or altruism vs. selfishness; no efficiency concerns. Does not capture harming in any other way than not sharing.
The senders will make choices in a series of binary games. The order of the games will be randomized. The games will be as follows: One of the options will be fixed at (10,10) in every game. The other option will change across games as follows: (11,13), (10,8), (9,6), (8,4), (7,2), (6,0).
The (11,13) option allows us to separate efficiency maximizers and altruists from egalitarians. The fewer (10,10) in the remaining games, the more willing to harm.
Measures: on top of inequality aversion, altruism and selfishness it adds efficiency concerns (Charness and Rabin 2002) and antisocial/spiteful behavior (Herrman et al., 2008).
The simple setting gives us an opportunity to manipulate the default option: either the (10,10) option is selected as a default or the alternative is chosen. The options are presented such that the selected option is presented at the left part of the screen. This option will again be useful in separating the role of moral costs from the role of social distance and reciprocity.
This section discusses the procedure of the experiment for the dictators. In the first round, there will be two parts. The first part will ask basic demographic questions and questions related to the experience of the individual with employment of domestic workers (if, and if so, for how long and of which nationality). It will also ask few general questions about the regulations related to employment of domestic workers. Lastly, it will ask a question about which two nationalities do the largest groups of domestic workers in their country have. The second part consists of the actual choices in dictator games. One of the treatments is implemented. The order of the games is randomized and the ordering is recorded.
The second round is different and its purpose is to determine whether the first round treatment has 1) persistent effects, 2) whether there the “first impressions” effect matters, and 3) whether it affects attitudes towards Filipinos in general. It starts immediately with choices in the same dictator games. The dictators are again matched with the same Filipina as in the first round. They will be told that “you are matched with the same person as last week”. After their dictator game choices, the dictators will be asked to take part in a single-target implicit association test testing their subconscious attitudes towards Filipinas (See Lowes et al. 2015; note that the set of photos will consist of photos of the Filipinas, including the matched one; although with lower precision, we will be able to elicit attitudes towards the specific individuals). To further elicit attitudes towards Filipinos in general, we will also ask a question related to social distance of the dictators and Filipinos inspired by a question on “views of Filipinos” as in Lowes et al. (2015): “In which of the following ways do you view Filipino people: very positively, somewhat positively, neutral, somewhat negatively or very negatively?”