This study compares risk preferences elicited from two different methods and the resulting inconsistency rates in response behavior. We also identify and compare how demographic and socioeconomic characteristics influence risk preferences elicited from the two methods. We use experimental and survey data collected from 332 randomly selected smallholder coffee farmers in Uganda. We find relatively low inconsistency rates in the response behavior and that both methods classify most farmers as risk averse. However, a closer inspection reveals significantly different risk results. Specific demographic and socioeconomic characteristics affect farmers’ risk preferences but are not stable across elicitation methods.