When deciding whether to invest in new skills, people must forecast how their own learning will unfold. I study how prior learning experiences shape these forecasts and the investment decisions that follow. In a controlled experiment, participants repeatedly attempt to acquire knowledge while I independently randomize two features of their environment: how likely each attempt is to succeed, and how much there is to learn. By eliciting their predictions about their own learning trajectories before and after this experience and comparing them to a Bayesian benchmark, I separate two kinds of belief error: parameter error, a misestimate of one's own learning ability, and function error, a misunderstanding of how that ability turns into skill over time. On average, experience reduces rather than improves forecast accuracy, and the effect comes entirely from negative experiences. After struggling, participants revise their predictions sharply downward, far beyond what the benchmark warrants. This over-revision does not come from updated beliefs about their ability. Instead, participants grow more confident in a misspecified model in which progress is mostly a matter of luck and individual ability matters little. The resulting pessimism is not justified by what they believe about their own ability, and it lowers investment in skill acquisition by 37%. These findings suggest that a central barrier to human capital investment may not be how accurately people assess their own ability, but how they understand the learning process itself.

Do discounts affect future demand through psychological mechanisms inconsistent with rational responses? We study this question in a field experiment in Malawi, whereby a direct sales team randomly assigned discounts for a preventive health care product, revisiting households a few weeks later to elicit their willingness to pay (WTP) for that and other goods. We document that past discounts significantly reduce WTP, with clear evidence of anchoring: the distribution of WTP spikes at round-1 prices, regardless of purchasing decisions. Strikingly, anchoring arises even for goods not offered at the first visit. Design choices allow us to rule out rational explanations for these patterns, such as intertemporal substitution, price-based inference about quality, or learning about costs. Results help reconcile conflicting findings from the literature and have broader implications for welfare analyses of price changes.

Perceptions of what potential partners value may shape major life choices, from career investments to the division of household labor. Yet little is known about whether these perceptions are accurate. We study this question using a lab-in-the-field experiment on a Swiss online dating platform, combining incentive-compatible elicitation of individuals’ preferences over partner characteristics with beliefs about what potential partners desire. We uncover large and systematic misperceptions. While revealed preferences are relatively progressive, beliefs remain strongly anchored in traditional gender norms. Most strikingly, women overestimate men’s preference for traditional domestic roles by roughly 500%. Men, in turn, substantially underestimate women’s valuation of egalitarian partners and incorrectly believe that high income can compensate for traditional gender attitudes. These misperceptions imply that both women and men may make suboptimal career and domestic investment decisions, potentially perpetuating gender inequality. In a follow-up field experiment, we test whether correcting these beliefs causally affects signaling behavior, match formation, and skill investment choices.