Replication of Study 1 by Mani, Mullainathan, Shafir, & Zhao (2013, Science)
Author
Vivian Huynh (vivhuynh@stanford.edu)
Published
December 15, 2023
Introduction
In the original study “Poverty Impedes Cognitive Function,” by Mani, Mullainathan, Shafir, and Zhao, researchers hypothesized that poverty has a direct effect on cognitive functions and conducted two studies to test their hypothesis. In these studies, it was evident that the cognitive capacities of low-income participants as opposed to high-income participants varied significantly when affected by financial situations. Mani, Mullainathan, Shafir, and Zhao claim that poverty itself “reduces cognitive capacity,” as a result of the financial stress that consumes cognitive resources, which further perpetuates poverty in the poor–financially, mentally, and behaviorally.
As an undergraduate in Symbolic Systems with a concentration in Cognitive Science, the idea of limited cognition as a result of poverty interests me because of the intersectionality of being a first generation, low-income student that studies the brain. Upon reading this article, skepticism definitely arose because being low-income or growing up in poverty does not necessarily imply that people in these circumstances will be less successful, however, there is truth and evidence that demonstrates how poverty structurally–and even generationally–affects the cognition of those living in poverty. There are many factors that impact low-income populations aside from finances, and this article provides evidence of the behaviors and thought processes as a result of limited cognition from poverty itself.
Description of Stimuli and Procedures
To replicate this experiment, we will conduct two studies: the first study will randomly assign low-income and high-income participants to an easy condition or hard condition, with the conditions varying from a low cost or high cost hypothetical financial scenario. Unlike the original study, we will be omitting the use of Raven’s Progressive Matrices Test to measure participants’ fluid intelligence, as the matrices may not be inclusive to populations of color, and have a questionable history (white-favoring, eugenics). Instead, we will follow through with examining non-verbal, analytical skills through a spatial compatibility test in each condition after evoking thoughts of financial concern in participants through financial scenarios. By the end of the experiment, we will gather data on the economic demographics of our participants, such as employment status, age, race, and socioeconomic status (SES).
Potential Challenges
Challenges that may arise include the ineffectiveness of hypothetical scenarios as they may not trigger financial concerns for low-income participants because the scenarios are not real. We also may not observe the same results as the original study due to the fact that our experiment will be hosted online, as opposed to in person, in a New Jersey mall. Additionally, there is no time limit or expected duration of the spatial compatibility test, so there might not be pressure for participants to perform as accurately as they can in the shortest amount of time possible. Because participants are able to take their time during this online experiment and do not have the external pressure of being proctored by researched in-person, their task accuracy may average higher than that of the original study. Cognition measures may be inaccurate due to external factors such as mental health, familial issues, and more, but may be excluded due to random assignment.
Original paper: https://github.com/psych251/mani2013_/blob/main/original_paper/science.1238041.pdf
OSF:https://osf.io/tx957
Methods
First, we will be using Prolific to recruit participants from different financial backgrounds, ideally ranging from $20,000 to $100,000. Although not specified in the original Mani et. al., 2013, study, we hope that our sample will have varying salaries in order to compare those in the lower end and the higher end. We will also filter participants between the ages of 25-80 years, as we want to filter out students or individuals that do not have a steady income. After recruiting participants, we will direct them to Cognition.Run to retrieve their consent to participate in the study, as well as to simulate the original experiment virtually.
When directed to Cognition.Run, participants will begin the experiment by reviewing hypothetical financial scenarios, then submitting survey responses on their decisions as to how they would budget a financial inconvenience to ensure that participants have understood and contemplated the given financial scenarios. By evoking thoughts of financial concern, we aim to observe the cognition of participants to analyze the effect of poverty on cognitive function. Participants will be randomly assigned either the easy condition or the hard condition of the financial scenario, with the easy condition containing less severe financial issues, and the hard condition containing more severe financial issues. Upon completing their responses to the scenarios, participants will be assigned a spatial compatibility task to measure participants’ cognitive capacities after some hypothetical hardships.
Power Analysis
The original study had an effect size of d = 0.88 with 95% power and a 0.05 alpha value. With that being said, the minimum sample size should contain at least 20 total participants in each condition (hard, easy).
Planned Sample
We hope to draw in participants of all gender identities, namely, male/female/non-binary, so long as they report their annual income, which will determine their SES background. The sample size of the original study was n = 101, meaning that 101 individuals participated. In our study, we will be recruiting 50 participants, where 25 subjects will be randomly assigned to the hard and easy condition. Although the original study does not provide an age range for participants the mean age was 35.3 years. We hope to recruit participants between 25-80 years old. With this age range, perhaps participants will have more financial security or a steady income, to rule out college students, or students without stable income. To aid the pre-selection of participants, we filtered out students on Prolific.
Materials
As noted in our Methods, participants will be recruited using Prolific where they will undergo a pre-screening to ensure that they are not actively students. Then, they will be redirected to Cognition.Run to complete a consent agreement and proceed to the experiment. Participants will be tasked to contemplate and answer hypothetical financial scenarios, which were retrieved from the original study in order to illicit similar procedures. Lastly, they will be assigned a non-verbal cognitive task (spatial compatibility) that involves pressing the ‘f’ and ‘j’ key when prompted by either a heart or a flower.
Procedure
“In experiment 1, participants (n = 101) were presented with four hypothetical scenarios a few minutes apart. Each scenario described a financial problem the participants might experience. For example: “Your car is having some trouble and requires $X to be fixed. You can pay in full, take a loan, or take a chance and forego the service at the moment… How would you go about making this decision?” These scenarios, by touching on monetary issues, are meant to trigger thoughts of the participant’s own finances. They are intended to bring to the forefront any nascent, easy to activate, financial concerns. After viewing each scenario, and while thinking about how they might go about solving the problem, participants performed two computerbased tasks used to measure cognitive function: Raven’s Progressive Matrices and a spatial compatibility task. The Raven’s test involves a sequence of shapes with one shape missing (27).
Participants must choose which of several alternatives best fits in the missing space. Raven’s test is a common component in IQ tests and is used to measure “fluid intelligence,” the capacity to think logically and solve problems in novel situations, independent of acquired knowledge (28, 29). The spatial incompatibility task requires participants to respond quickly and often contrary to their initial impulse. Presented with figures on the screen, they must press the same side in response to some stimuli but press the opposite side in response to others. The speed and accuracy of response measures cognitive control (30), the ability to guide thought and action in accordance with internal goals (31). Both are nonverbal tasks, intended to minimize the potential impact of literacy skills. Upon completion of these tasks, participants responded to the original scenario by typing their answers on the computer or speaking to a tape recorder and then moved on to the next scenario (an analysis of participants’ responses to the scenarios is available in table S1). We also collected participants’ income information at the end of the experiment. Participants were randomly assigned either to a “hard” condition, in which the scenarios involved costs that were relatively high (for example, the car would require $1500 to fix); or to an “easy” condition, where costs were lower (for example, the car would require $150 to fix). Because the sums in the easy condition are small, we expected this condition to evoke few of one’s own monetary concerns, for either poor or rich participants. In contrast, the large sums in the hard condition, we hypothesized, would evoke monetary concerns in the poor but not in the rich participants.”
Our procedure is similar to that of the original study, as it will implement similar, if not same, hypothetical financial scenarios followed by one cognitive function task, the one being spatial compatibility. The difference is that we will be conducting the experiment remotely, on Cognition.Run. Here, participants will demonstrate their understanding and contemplation of the prompt by providing a short answer to the hypothetical scenarios. We will not be giving participants the Raven’s Matrices tasks, but instead narrow down on the spatial compatibility task. Like the original study, there will be three trials for the spatial compatibility task, and there will be three blocks of these trials. Having the same number of trials will result in similar accuracy ranges for the task, as scores will be in buckets of 0%, 33%, 67%, and 100%.
Analysis Plan
To analyze our data, we want to compare the overall accuracy of the spatial compatibility task that measures cognitive function, between high income and low income groups. It would be most visually helpful to produce a bar chart to make the comparison. We also want to analyze the linear regression model for participants in both the hard and easy condition to see the effect poverty has on cognitive function.
In our first course of action, we will recruit a total of 50 participants, randomly assigning 25 and 25 to the hard and easy conditions. Then, we will use the U.S. Federal Poverty guidelines to assign each participant to their corresponding SES status. Because Prolific users complete many experiments over the course of the day, we will have participants provide a short answer to the hypothetical scenarios to ensure that they have read and understood the prompts. After their responses, we will administer a practice round of ten trials for the spatial compatibility task before participants complete the test trials. We will measure the duration of these trials, as well as their accuracy. Although not noted in the original study, it is suspected that perhaps the duration of time spent on the trials may play a role on accuracy. In other words, a longer duration may procure more accurate results, and a shorter duration may procure less accuracy.
Differences from Original Study
In this study, we understand that using an online platform may affect the results as a whole since not everyone who is low-income or high-income may have the time to participate in this study, let alone be exposed to it. Moreover, since the study is conducted online, we understand that the experience of participants will not be the same as the original study that was conducted in a mall in New Jersey. Participants on Prolific know that they will be participating in experiments and may be mentally primed to perform, whereas mall shoppers may have been caught by surprise by the introduction of the study and may have experienced external pressure of participating in an experiment that was conducted in-person. It is notable that we will be using Cognition.Run to gather survey responses, which may or may not encourage falsity in terms of income, meaning, there is no way to confirm the income of each participant, apart from taking their word.
Methods Addendum (Post Data Collection)
Actual Sample
In our experiment, we were able to recruit 50 participants through Prolific that were all approved for finishing the experiment tasks. There was an even distribution between sex, the mean income was $62,946, and the mean age was 42 years of age. A noteworthy comment would be that most Prolific particpants were white.
Differences from pre-data collection methods plan
In the original study, researched conducted three trials in their spatial compatibility task, and we did the same. We kept the same number of trials because we wanted the accuracy scores to match the same buckets as the original study. Conversely, we had three blocks of three trials to see if there were differences among blocks.
Results
Data preparation
We want to read the dataframe from our Cognition.Run responses, as well as the scores received from the cognition tasks. First, we have to tidy the data and organize participants’ demographics including: age, gender identity, income, and so on. It would be valuable to do this as it will allow us to potentially make inferences about each social aspect. We will not include participants that do not answer all of the required questions or do not provide consent to participate in the experiment on Cognition.Run.
### Data Preparation#### Load Relevant Libraries and Functionslibrary(tidyverse)
── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
✔ dplyr 1.1.3 ✔ readr 2.1.4
✔ forcats 1.0.0 ✔ stringr 1.5.0
✔ ggplot2 3.4.3 ✔ tibble 3.2.1
✔ lubridate 1.9.3 ✔ tidyr 1.3.0
✔ purrr 1.0.2
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag() masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
# remove JSON text from survey responsesdemo_data$Age <-gsub('"', '', as.character(demo_data$Age))demo_data$Race <-gsub('"Q1:"', '', demo_data$Race)demo_data$Race <-gsub('"', '', demo_data$Race)demo_data$Gender <-gsub('"Q2:"', '', demo_data$Gender)# updating Gender Columndemo_data$Gender <-gsub("[Mm]ale", "M", demo_data$Gender)demo_data$Gender <-gsub("Man", "M", demo_data$Gender)demo_data$Gender <-gsub("Woman", "F", demo_data$Gender)demo_data$Gender <-gsub("[Ff]emale", "F", demo_data$Gender)demo_data$Gender <-gsub("[Ff]eM", "F", demo_data$Gender)demo_data$Gender <-gsub("femme", "F", demo_data$Gender)demo_data$Gender <-gsub("non binary", "NB", demo_data$Gender)demo_data$Gender <-gsub('"Q3":', "", demo_data$Gender)# add demo data to main dataframedata_filtered <-merge(data_filtered, demo_data, by =c('participant_id')) ########### CLEANING DATA ############## this separates data by block per participantclean_data <- data_filtered %>%filter(task =="response") %>%group_by(participant_id, group, ses_group, block, income, Age, Race, Gender) %>%summarize(mean_acc =mean(as.numeric(accuracy)),mean_rt =mean(as.numeric(rt)),acc_sd =sd(as.numeric(accuracy)),acc_n_obs =length(as.numeric(accuracy)),acc_sem = acc_sd /sqrt(acc_n_obs),acc_ci = acc_sem *1.96,rt_sd =sd(as.numeric(rt)),rt_n_obs =length(as.numeric(rt)),rt_sem = rt_sd /sqrt(rt_n_obs),rt_ci = rt_sem *1.96 )
`summarise()` has grouped output by 'participant_id', 'group', 'ses_group',
'block', 'income', 'Age', 'Race'. You can override using the `.groups`
argument.
# this has the overall average performance between all trials for each participantcleaner_data <- data_filtered %>%filter(task =="response") %>%group_by(participant_id, group, ses_group, income, Age, Gender) %>%summarize(mean_acc =mean(as.numeric(accuracy)),mean_rt =mean(as.numeric(rt)),acc_sd =sd(as.numeric(accuracy)),acc_n_obs =length(as.numeric(accuracy)),acc_sem = acc_sd /sqrt(acc_n_obs),acc_ci = acc_sem *1.96,rt_sd =sd(as.numeric(rt)),rt_n_obs =length(as.numeric(rt)),rt_sem = rt_sd /sqrt(rt_n_obs),rt_ci = rt_sem *1.96 )
`summarise()` has grouped output by 'participant_id', 'group', 'ses_group',
'income', 'Age'. You can override using the `.groups` argument.
# change age and income to numericcleaner_data$Age <-as.numeric(cleaner_data$Age)cleaner_data$income <-as.numeric(cleaner_data$income)####################################################### TASK ANALYSIS ######################################################### ######### DEMOGRAPHIC ANALYSIS ############age_graph <-ggplot(cleaner_data, aes(x = Age)) +geom_bar() +scale_y_continuous(breaks =seq(0, 10, 1)) +ggtitle("Distribution by age") +ylab("# of Participants")age_graph
# What is the average age? 42.45 (rounded)mean(cleaner_data$Age, na.rm =TRUE) # some did not report age
[1] 43.29091
income_graph <-ggplot(cleaner_data, aes(x = income)) +geom_bar() +ggtitle("Income Distribution") +ylab("# of Participants") +xlab("Approximate Total Income (thousands)")income_graph
# What is the average income? 62946.43 mean(cleaner_data$income) # no NA since they were removed
[1] 62636.36
# What is the total gender breakdown? gender_table <- demo_data %>%group_by(Gender) %>%summarise(total =n())######### OVERALL ANALYSIS ############# this is to create Figure 1: comparing between SES and conditionstask_perf <- cleaner_data %>%group_by(group, ses_group) %>%summarize(accuracy =mean(mean_acc), na.rm =TRUE,rt =mean(mean_rt),acc_sd =sd(mean_acc),acc_n_obs =length(participant_id),acc_sem = acc_sd /sqrt(acc_n_obs),acc_ci = acc_sem *1.96,rt_sd =sd(mean_rt),rt_n_obs =length(participant_id),rt_sem = rt_sd /sqrt(rt_n_obs),rt_ci = rt_sem *1.96,mean_age =mean(as.numeric(Age), na.rm =TRUE))
`summarise()` has grouped output by 'group'. You can override using the
`.groups` argument.
`summarise()` has grouped output by 'group', 'ses_group'. You can override
using the `.groups` argument.
Confirmatory analysis
Our analysis will be successful if we able to reproduce the same results in the study, where there are no significant differences between participants in the easy condition based on our t-test, as well as if the t-test demonstrates that low SES participants in the hard condition will have lower accuracy. Moreover, we will be successful if our two-way analysis of variance (ANOVA) demonstrates a robust interaction between income and condition. Our linear regression models have to confirm that income will be predictive of accuracy for participants in the hard condition only, as Mani et al., (2013), claims that higher-income individuals perform well in the hard condition.
# Create Figure 2 - Accuracy Regression Chartfigure2 <-ggplot(cleaner_data, aes(x = income, y = mean_acc,color = group)) +geom_point() +scale_color_manual(values=c("#000000", "#999999")) +geom_smooth(method=lm, aes(fill = group)) +ggtitle("Figure 2: Accuracy by Income and Group") +labs(y ="Mean Accuracy", x ="Income (in thousands)")figure2
`geom_smooth()` using formula = 'y ~ x'
NOTE: We have encountered an issue where the mean accuracy exceeds 1, which is strange, considering our printed mean_acc values are all less than 1. We suspect that this may be an issue due to rounding, but are unsure. Figure 2 displayed above is not an accurate model.
################################################### STATISTICAL ANALYSIS #################################################### ####### ANALYSIS OF OVERALL DATA ########### TWO-WAY ANOVAaccuracy_aov <-aov(mean_acc ~ group * ses_group, cleaner_data)summary(accuracy_aov)
Df Sum Sq Mean Sq F value Pr(>F)
group 1 0.00063 0.000628 0.165 0.686
ses_group 1 0.00083 0.000829 0.218 0.643
group:ses_group 1 0.00297 0.002967 0.778 0.382
Residuals 51 0.19445 0.003813
### Regression Analysisaccuracy_model <-lm(mean_acc ~ group * ses_group, cleaner_data)summary(accuracy_model)
Call:
lm(formula = mean_acc ~ group * ses_group, data = cleaner_data)
Residuals:
Min 1Q Median 3Q Max
-0.18518 -0.06957 0.03704 0.03865 0.04444
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.955556 0.027615 34.603 <2e-16 ***
groupHARD 0.044444 0.045094 0.986 0.329
ses_groupRICH 0.005797 0.030469 0.190 0.850
groupHARD:ses_groupRICH -0.042834 0.048561 -0.882 0.382
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 0.06175 on 51 degrees of freedom
Multiple R-squared: 0.02225, Adjusted R-squared: -0.03527
F-statistic: 0.3868 on 3 and 51 DF, p-value: 0.763
To confirm that there are differences between low-income and high-income groups, we will be using a t-test to measure the differences of cognitive capacity scores, hoping to see variation among scores depending on the participants’ income. The t-test will allow us to confirm whether or not there is significance in our result against the null hypothesis. Mani et. al., 2013, claims that low-income populations are cognitively impacted by financial stress, as opposed to high-income populations. To test this hypothesis, researchers assigned cognitive capacity tests and well as a spatial compatibility task to compare the differences in scores.
t.test(mean_acc ~ ses_group, data = cleaner_data)
Welch Two Sample t-test
data: mean_acc by ses_group
t = 0.49392, df = 10.856, p-value = 0.6312
alternative hypothesis: true difference in means between group POOR and group RICH is not equal to 0
95 percent confidence interval:
-0.03479740 0.05489196
sample estimates:
mean in group POOR mean in group RICH
0.9722222 0.9621749
boxplot(mean_acc ~ ses_group, data = cleaner_data)
Exploratory analyses
Any follow-up analyses desired (not required).
# Create Figure 3 - RT Bar Chartfigure3 <-ggplot(task_perf, aes(x = ses_group, y = rt, fill = group)) +geom_bar(position="dodge", stat="identity") +geom_errorbar(aes(ymin = rt - rt_ci, ymax = rt + rt_ci), width=.2,position=position_dodge(.9)) +ggtitle("Figure 3: Reaction Time by Group") +ylab("RT (ms)") +xlab("SOCIOECONOMIC STATUS") +scale_fill_manual(values=c("#D3D3D3", "#999999"))figure3
### TWO-WAY ANOVArt_aov <-aov(mean_rt ~ group * ses_group, cleaner_data)summary(rt_aov)
Df Sum Sq Mean Sq F value Pr(>F)
group 1 1605 1605 0.031 0.860
ses_group 1 56359 56359 1.099 0.299
group:ses_group 1 76 76 0.001 0.969
Residuals 51 2615335 51281
figure4 <-ggplot(cleaner_data, aes(x =as.numeric(cleaner_data$income), y =as.numeric(mean_rt),color = group)) +geom_point() +geom_jitter() +scale_color_manual(values=c("#000000", "#999999")) +geom_smooth(method=lm, aes(fill = group)) +ggtitle("Figure 4: Reaction Time by Income and Group") +labs(y ="Mean Reaction Time (ms)", x ="Income (in thousands)")figure4
`geom_smooth()` using formula = 'y ~ x'
rt_model <-lm(mean_rt ~ group * ses_group, cleaner_data)summary(rt_model)
Call:
lm(formula = mean_rt ~ group * ses_group, data = cleaner_data)
Residuals:
Min 1Q Median 3Q Max
-339.42 -161.17 -27.14 97.16 512.91
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 693.480 101.273 6.848 9.47e-09 ***
groupHARD -22.884 165.378 -0.138 0.890
ses_groupRICH 88.511 111.740 0.792 0.432
groupHARD:ses_groupRICH 6.869 178.091 0.039 0.969
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 226.5 on 51 degrees of freedom
Multiple R-squared: 0.02171, Adjusted R-squared: -0.03584
F-statistic: 0.3773 on 3 and 51 DF, p-value: 0.7698
accuracy_block_model <-lm(mean_acc ~ group + ses_group + block, clean_data)summary(accuracy_block_model)
Call:
lm(formula = mean_acc ~ group + ses_group + block, data = clean_data)
Residuals:
Min 1Q Median 3Q Max
-0.31789 0.01900 0.03415 0.04574 0.05681
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.999710 0.029777 33.573 <2e-16 ***
groupHARD 0.007507 0.016321 0.460 0.646
ses_groupRICH -0.011066 0.023142 -0.478 0.633
block -0.015152 0.009947 -1.523 0.130
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 0.1043 on 161 degrees of freedom
Multiple R-squared: 0.01663, Adjusted R-squared: -0.001697
F-statistic: 0.9074 on 3 and 161 DF, p-value: 0.4389
# Create Figure 6 - Reaction Time by Blockfigure6 <-ggplot(block_perf, aes(x = ses_group, y = rt, fill = group)) +geom_bar(position="dodge", stat="identity") +geom_errorbar(aes(ymin = rt - rt_ci, ymax = rt + rt_ci), width=.2,position=position_dodge(.9)) +facet_wrap(~ block) +ggtitle("Figure 6: Reaction Time by Block") +ylab("RT (ms)") +xlab("SOCIOECONOMIC STATUS")figure6
# regression analysis for block, group, income on RTrt_block_model <-lm(mean_rt ~ group + ses_group + block, clean_data)summary(rt_block_model)
Call:
lm(formula = mean_rt ~ group + ses_group + block, data = clean_data)
Residuals:
Min 1Q Median 3Q Max
-387.14 -192.46 -26.12 128.07 1361.03
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 788.90 76.24 10.347 <2e-16 ***
groupHARD -16.96 41.79 -0.406 0.685
ses_groupRICH 91.21 59.26 1.539 0.126
block -48.82 25.47 -1.917 0.057 .
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 267.1 on 161 degrees of freedom
Multiple R-squared: 0.03657, Adjusted R-squared: 0.01862
F-statistic: 2.037 on 3 and 161 DF, p-value: 0.1108
Discussion
Summary of Replication Attempt
In the original study “Poverty Impedes Cognitive Function,” by Mani et al., (2013), researchers reported that a two-way analysis of variance (ANOVA) “revealed a robust interaction between income and condition”, with a p-value of < .001 for performance on the cognitive control task. In our attempt to replicate, our two-way ANOVA revealed that there was no significant interaction between income and group [F(1,50) = 0.565, P = 0.456]. Figures 1 and 2 also indicate that were no differences would between groups through both a t-test or a linear regression analysis. Instead, in our experiment, we actually observed that the high-income group had a smaller accuracy in both conditions, contrary to performing well in the hard condition, as demonstrated in Mani et al., (2013). We also want to reiterate that there is an error in Figure 2, as the mean accuracy exceeds 1, despite our printing of the variable showcasing no values greater than 1. In conclusion, we failed to replicate the results revealed in the original study.
Commentary
We suspect that one reason why the overall mean accuracy of both the low-income and high-income group was high was because we gave participants ten practice trials before proceeding with the experiment. Because of the long practice trial, participants may have had an advantage when it came to the three test trials. This may not be the case for the overall high accuracy because in the original study, researchers also have their participants ten practice trials. Contrary to the original study, our experiment was conducted remotely, rather than in person at the mall.
In the original replication conducted by Gabriel Reyes, Reyes noted that their results may have been skewed due to the fact that most of the participants in the study were college students, or did not have stable income. As a result, in our study, we added a minimum of 25 years in order to recruit participants that could potentially have steady income. From our exploratory analyses, it is revealed that very few participants answered incorrectly for a trial in a single block. We can also see that the reaction time of the high-income group was a few seconds larger than that of the low-income group. This could be due to age or the interpretation of directions, or even the time spent curating a 1-2 sentence response to the hypothetical financial scenarios.
