This study focuses on the impact of macroeconomic factors within credit risk assessments in the peer to peer lending market. The credit risk industry faces a classification problem, i.e. predicting at initiation whether a loan will default or not. Existing literature obtained statistically significant logistic regression coefficients for macroeconomic variables. The predictive capability in terms of classification performance of macroeconomic variables in machine learning-based models has, however, not been tested. Besides, previous literature on explaining black-box algorithms in the credit risk industry mostly focused on the prediction fidelity of XAI tools and not on the economic interpretation of the SHAP
values. This research attempts to improve state-of-the-art classification performance by including macroeconomic variables in the decision-making process. In addition, the importance and economic interpretation of the macroeconomic features are assessed by analyzing the SHAP values. The study uses the LendingClub loan data set from 2012 up to 2020. The main results include that machine learning algorithms are capable of leveraging macroeconomic features to improve overall classification performance. Furthermore, there appears to be a selection effect present in credit risk assessment, i.e. during economic downturns, default probabilities of originated loans decrease due to investors’ demand for stricter lending conditions.