Suppose you had to find the effect of seatbelt rules on road accident fatalities. Would you choose to run a randomized experiment or would it make sense to use natural experiments​ here? Explain.