We administer a newly-designed survey to a large panel of retail investors who have substantial wealth invested in financial markets. The survey elicits beliefs that are crucial for macroeconomics and finance, and matches respondents with administrative data on their portfolio composition and their trading activity. We establish five facts in this data: (1) Beliefs are reflected in portfolio allocations. The sensitivity of portfolios to beliefs is small on average, but varies significantly with investor wealth, attention, trading frequency, and confidence. (2) It is hard to predict when investors trade, but conditional on trading, belief changes affect both the direction and the magnitude of trades. (3) Beliefs are mostly characterized by large and persistent individual heterogeneity; demographic characteristics explain only a small part of why some individuals are optimistic and some are pessimistic. (4) Investors who expect higher cash flow growth also expect higher returns and lower long-term price-dividend ratios. (5) Expected returns and the subjective probability of rare disasters are negatively related, both within and across investors. These five facts challenge the rational expectation framework for macro-finance, and provide important guidance for the design of behavioral models.