MLM Disclosure Experiment

Austin Miller

Introduction

Here you can access the documents I created to present and discuss the R scripts that I wrote to analyze the data generated by a recent Amazon Mechanical Turk survey experiment. The goal of the analysis is to assess the impact of voluntary income disclosures in MLM marketing materials on consumer interest and earnings expectations. All participants were introduced to an MLM opportunity using marketing materials from the website of an actual MLM firm. The control group did not receive any income disclosure information; treatment group 1 received the income disclosure document created by the MLM firm itself; and treatment group 2 received an augmented form of the firm’s income disclosure information that included a graph and presented how many participants in the firm actually earned zero dollars. The analysis from this experiment was recently published for publication as “Income Disclosure and Consumer Judgment in a Multi-level Marketing Experiment” in Journal of Consumer Affairs.

This document also employs R Shiny apps,

described here: Heterogeneous Treatment App

and here: Logarithm App

and here: correlations app

Analysis

MLM Experiment- Analysis

This script contains all of the regression analysis that supports the final paper. This includes:

• Main Model of the Income Disclosure Effects on All Dependent Variables
• A Graphical Justification for Logarithmic Transforms of Dependent Variables
• Alternative Interest Specifications
• Models of Heterogeneous Treatment Effects
• Models of the Effects on Interest of Thinking About Earnings

Data Cleaning

MLM Experiment- Data Cleaning

This script cleans the raw data and prepares it for analysis. This document includes:

• a description of all data transformations
• a discussion about possible errors in the responses
• a description of dropped observations

Raw and clean data files are available here: github.com/milleroztn/MLMExperiment/tree/main/data

Graphs

MLM Experiment- Graphs

The next script generates and discusses graphs from the data, including:

• Distributions of Key Variables, by Treatment Group
• Correlations of Interest v. Earnings
• Correlations of Better-than-average Effect with Interest and Earnings

Graph files are available here: github.com/milleroztn/MLMExperiment/tree/main/graphs

Summary Tables

MLM Experiment- Summary Tables

This script generates and discusses summary tables from the data, including:

• Summary Statistics for All Variables
• Significant Correlations with Dependent Variables
• Distributions of Dependent Variables Across Treatment Groups, including tests for differences in means, medians, and variances

The generated *.csv files from this script are available here: github.com/milleroztn/MLMExperiment/tree/main/descriptive