* Title: Does banknote quality affect counterfeit detection? Experimental evidence from Germany and the Netherlands ? * Authors: Frank van der Horst, Martina Eschelbach, Susann Sieber, Jelle Miedema ****************************************************************************** ****************************************************************************** * Information on dataset "counterfeits.asci" ****************************************************************************** ****************************************************************************** *** Number of observations: 511 (consumers and cashiers from Germany and the Netherlands) *** Source: Tests and interviews on the recognition of counterfeits conducted by the Deutsche Bundesbank and De Nederlandsche Bank (June 2014 and August 2015) *** Variables: * id = personal id (within each country) * nl = 1 if Dutch, 0 if German * professional = 1 if cashier, 0 if consumer * age = age in years * female = 1 if female, 0 if male * educ1 = 1 if secondary education, 0 otherwise * educ2 = 1 if higher secondary education, 0 otherwise * educ3 = 1 if university degree, 0 otherwise * educ4 = 1 if no secondary education or education missing, 0 otherwise * handicaps = 1 if visual handicaps during the tests, 0 otherwise * payment1 = 1 if preferred payment mode is cash, 0 otherwise * payment2 = 1 if preferred payment mode is card, 0 otherwise * payment3 = 1 if cash and card are equally preferred, 0 otherwise * checked = 1 if individual has checked banknotes in the last 6 months, 0 otherwise * clean = 1 if clean test set during test, 0 otherwise * nb_selected = number of banknotes declared as counterfeits * correct_counterfeits = number of counterfeits correctly identified * correct_genuines = number of genuine banknotes correctly identified * time_stack_1 = sorting time stack 1 in minutes * time_stack_2 = sorting time stack 2 in minutes * time_stack_3 = sorting time stack 3 in minutes * time_stack_4 = sorting time stack 4 in minutes * time_stack_5 = sorting time stack 5 in minutes * set = id of test set * item_hits = number of security features checked ****************************************************************************** ****************************************************************************** * Empirical analysis using Stata: ****************************************************************************** ****************************************************************************** clear clear matrix clear mata version 12 set more off import delimited using counterfeits.asci, clear ****************************************************************************** * Calculation of performance measures: ****************************************************************************** * Hit rate generate hit_rate=correct_counterfeits/20 * FA_rate generate FA_rate=(nb_selected - correct_counterfeits)/180 * A_prime generate hilf1=min(hit_rate, 0.975) generate hilf2=max(FA_rate, 0.025) generate A_prime=. replace A_prime=0.5+((hit_rate-FA_rate)*(1+hit_rate-FA_rate)/(4*hit_rate*(1-FA_rate))) if hit_rate>=FA_rate replace A_prime=0.5-((hit_rate-FA_rate)*(1+hit_rate-FA_rate)/(4*hit_rate*(1-FA_rate))) if hit_rate