Because of its relation to economic growth, there is a policy interest in mea- suring social capital and average trust as its currently most important proxy. Thereby a main focus is determining its variation across groups with different individual characteristics. In this paper we combine the virtue of laboratory experiments and survey data analysis. We present results from a novel experiment conducted on a representative sample of the Dutch population. The advantages of this combination of methods are to shed light on four almost undocumented yet important issues in trust economics. Our results can briefly be summarized as follows. We do not find evidence of a participation selectivity bias which is a serious concern for laboratory experiments which rely almost exclusively on volunteer participants. Contrary to the existing literature, we find that stated trust measures correlate with experimental trust. The differences in parameter estimates across both measures are significant, but do not hold a jointly test. We also find that the age and education profiles of trust are opposite to those of reciprocal behavior. Finally, we find that the choice of proxy variable for social capital matters greatly, leading to very different inferences.