Keynes (1936) suggested that average expectations matter for the expectations formed by individuals. Recent contributions have examined the theoretical implications for asset pricing. In this paper, we present first empirical tests of this form of expectation formation. Using data from a large survey of professional forecasters, we find strong and robust evidence that the average of expected forecasts matter much for individual forecasters' expectations about stock market movements. We evaluate what kind of forecasters are more prone to be influenced by higher-order expectations. We find that young forecasters, forecasters who rely more on macroeconomic variables when forming their expectations, and forecasters who are less (over-)confident rely more on average expectations.