Identifying and Addressing Structural Racism

2023-06-01

2023-12-31

perceptions of racial discrimination, reported healthcare demand, healthcare take-up

We hypothesize that in the presence of structural racism, a negative signal about one system will influence outcomes in the other. Our first experiment will elicit priors on experiences in the criminal justice system and the healthcare system, with a focus on understanding participant willingness to take up healthcare. We will then randomize respondents to either a treatment or control group. The treatment group will receive a series of statistics and a narrative on racial inequality in the criminal justice system, and the passive control will receive statistics and a narrative about lightning strikes (i.e. a placebo). The narrative vignettes related to discrimination in the criminal justice system will be informed by details of actual cases of discrimination and will be produced by the journalism production company The Invisible Institute.

After making salient racial disparities in the criminal justice system, we will again measure perceptions of racial discrimination in both the healthcare and criminal justice systems as well as other systems such as financial markets and education. In the absence of structural racism, and/or the absence of perceived structural racism, individuals in the first treatment group may only update regarding criminal justice; however, if the two systems are “correlated” due to structural racism and if individuals internalize this correlation, i.e., if a signal in one system is informative about the distribution of outcomes of another, then making discrimination in the criminal justice system more salient will have spillover effects on perceptions of the healthcare system. Individuals in this group may then be less willing to seek medical care.

In the second experiment, we will test the hypothesis that the updating process observed in Experiment 1 can be weakened by making salient the actions of medical professionals who denounce racial discrimination. In this experiment, individuals will again be randomized into one of four groups: first, they will be assigned to one of the two experimental arms from our first experiment, and then separately randomized into a second treatment: in this second treatment, the treatment group will be shown images and a narrative video of doctors protesting against police violence in the wake of the death of George Floyd. The control group will be shown unrelated images of doctors protesting climate change. We will then measure the effect of this second treatment on healthcare demand. We will measure self-reported health demand as in Experiment 1. We will also measure actual take-up of a coupon for telehealth services from a third-party provider, Plush Care. This telehealth provider is willing to create a signal of solidarity specific to their firm, linking the treatment to the same healthcare provider to which participants will be given subsidized access.

Not available

Randomization will be conducted using a computer random number generator.

We will randomize at the level of the individual.

No

N/A

Our target sample size is 1,000 individuals for our first experiment and 7,500 individuals for our second experiment. Reflecting the focus of our study on communities affected by racial discrimination, 80% of our respondents in Experiment 1 will be non-Hispanic Black, while 20% of the respondents will be non-Hispanic White. All respondents in Experiment 2 will be non-Hispanic Black. As we are particularly interested in examining marginalized Americans who may be reluctant to access healthcare services due to structural racism, we will restrict our sample in both experiments to individuals who do not have a regular primary care physician. We will require all participants to be age 18 or older and currently live in the United States. To ensure our sample is representative of this population, we will institute race-specific age, gender, education, region, and income quotas during recruitment.

In Experiment 1, 500 individuals will be assigned to the treatment and control groups, respectively. In Experiment 2, 3,750 individuals will first be randomized to each of the two experimental arms from our first experiment, and then separately randomized into a second group: 1,875 individuals will be randomized to the second treatment group and 1,875 individuals will be randomized to the second control group.

In a recent project by Alsan and Eichmeyer (in progress), the authors leverage an experiment to identify the most persuasive ways to encourage low socioeconomic males living in the U.S. to take up seasonal flu vaccination. We use the means and standard deviations for the control group post-intervention across three outcomes from this study as a basis for our power calculations, as our experiment will also focus on low socioeconomic respondents and features healthcare demand as an outcome. We find that if there are underlying effect sizes of 8.5 percentage points for Experiment 1 and 5.9 percentage points for the key interaction term in Experiment 2, we should be able to detect a significant difference in outcomes. Based on these MDEs, we propose recruiting a sample of 1,000 participants for the first experiment and 7,500 participants for the second experiment. We believe that the MDEs are realistic given the economics literature on nudges and medical care take-up. For example, Alsan et al. (2019) study the effect of a race concordant health provider on the demand for preventive care among African American men and find that participants assigned to a Black doctor are 18 percentage points more likely to take up preventive care relative to those assigned to non-Black doctors. Similarly, Altmann and Traxler (2014) randomly provide patients with reminder messages about dental health prevention and find that individuals who receive a reminder are roughly 10 percentage points more likely to schedule a check-up, on average.

AEA_PAP_feb2022.pdf