Causal Inference

Designing successful Kickstarter Crowdfunding campaigns: A causal inference approach

Crowdfunding has become a straightforward and efficient way to raise money. We use causal inference to test how three campaign design choices impact success.

technologies

R (tidyverse, RItools, optmatch)

methods

Observational study, covariate matching; Fisher permutation test, Neymanian Hypothesis test

results

Established causal link improving campaign success from three Kickstarter features

Project overview

Using data collected and processed from the kickstarter website, we analyse whether any of prior experience, a feature tag called "projects we love", and allowing a tiny ($1) minimum reward, increase chance of a campaign reaching its funding target.

Execution

Given the data was collected retrospectively this was an observational study. To justify strong ignorability ee performed propensity score matching on the covariates, and run hypothesis tests on the Average Treatment Effects.

Results

For each of the three Kickstarter design features, we found very strong evidence that they caused an increase in the chance of a campaign reaching its funding goal. Sensitivity analysis suggests more covariates could bolster the causal claim.