Reversal and magnitude effects in long-term time preferences: Results from a field experiment