Is the cross sectional data at each date available in time to use as a prediction?


During the submission are we to assume we will be able to INCLUDE the cross-sectional data that reveals itself at the same time as the returns? I am trying to figure out if during my testing I should be predicting y at t + 1 with data from X at time t -n through time t or if I can include the X values at time t + 1 in the prediction.



You are predicting y_t for some t, using a function f(x_t). This means that when you are asked to predict y_t you have access to all the x_0, x_1, …, x_t, including the x_t; you moreover have access to all the y_0, y_1, …, y_{t-1}.

You do not have access to y_t and following targets, not future states of X_{t+n}.

Hope this clarifies.