We want to study how the availability of financial resources impacts individual asset holdings. We will exploit state lotteries in the United States to be able to tackle this question causally. This is the proposed regression model:

assets = 0 + 1lottery + 2educ + 3age + 4female + u (1) where assets denotes the level of asset holdings in USD, lottery is a dummy variable equal to one in all the periods after an individual won the lottery and equal to zero otherwise, educ is years of education, age is age in years and female is a dummy variable taking value one if the individual is a woman and zero otherwise. Assuming that all of these contain the relevant variables, which is the best data-set to answer the research question at hand? (a) A cross-sectional data-set with information about a sample of individuals who won and who did not win the lottery, covering all states with state lotteries. (b) A longitudinal data-set with information about a sample of individuals who won and who did not win the lottery for at least one period before and one period after they won, covering the whole United States. (c) Two repeated cross-sections of individuals after they won the lottery, covering the whole United States. (d) A panel data-set with information about a sample of individuals after they the won lottery for at least two periods after they won, covering all states with state lotteries.