Value of the data •These data are useful for understanding how individual characteristics (such as moral foundations), opinions, and situational circumstances predict observed (rather than self-reported) other-regarding behavior, specifically one's willingness to donate money to a predefined set of charities. Furthermore, the randomized assignment of charity sets allows researchers to investigate how these relationships are affected by the specific charities presented to each respondent. •Researchers interested in the determinants of charitable giving as well as researchers performing meta-analyses related to other-regarding behavior in economic games, the Moral Foundations Questionnaire, and Amazon Mechanical Turk data may all benefit from these data. This dataset also allows researcher to directly replicate the original study associated with this dataset. •The dataset not only includes donation decisions, but also a respondent's charity preference. The determinants of charity preference were not explored in the original paper associated with this dataset. There are also many potential determinants of charitable giving in the dataset that have yet to be explored (e.g. engagement in environmental issues). •Because respondents were required to be MTurk Masters workers and were paid a relatively high wage ($0.50 for approximately 5–10 minutes of work) this dataset offers a useful point of comparison for similar studies that do not use MTurk workers or ones that use MTurk workers but do not take the same measures to ensure data quality. Data quality can be measured by performance on attention checks in the dataset (some of which are part of the standard MFQ) as well as the internal consistencies of the responses to the MFQ questions. Furthermore, detailed raw data from Amazon Mechanical Turk can allow researchers to investigate whether responses and characteristics of respondents differ by the characteristics of the work request to better understand the external validity of Mechanical Turk data. •The dataset is large enough to allow various partitions to investigate heterogeneous relationships in the data without losing substantial statistical power.