The perception and integration of sequential numerical information is a common cognitive task. It is a prerequisite for experience-based economic choices, but it is usually not part of economic decision theory. To better understand the process of symbolic number integration and its influence on economic behavior, we performed three experimental studies that examined mean estimates and economic valuations of continuous number distributions. The results indicate that participants valued random number distributions below their respective arithmetic means and valued distributions as lower when their variance increased, indicating risk aversion. A similar though less pronounced pattern also occurred in the matched mean estimation task where accuracy was incentivized and preferences played no role. These patterns suggest that seemingly risk-averse preferences are partly due to cognitive biases when perceiving and estimating numbers. In addition, participants apparent economic preference for right-skewed outcome distributions could be attributed mainly to estimation biases. We discuss the extent to which the results can be explained based on a compressed mental number line and different sample weighting models. Finally, a new model that can account for the qualitative data pattern and has stronger overweighting of lower than higher numbers as its core feature is developed. Together, our results indicate that basic cognitive processes in perceiving and integrating number sequences play a key role in understanding experience-based economic behavior.