Table 1 Definitions and coding of the concepts in the bi-variate and logistic regressions
Dependent variable | |
FRAUD, as defined by the FraudeHelpDesk | 1) A CEO-fraud occurs when an employee receives an e-mail or a phone call from someone impersonating the Chief Executive Officer (CEO) or the Chief Financial Officer (CFO) to transfer a substantial amount of money to a foreign bank accounta 2) Fraudulent contracts: advertising agencies approach entrepreneurs offering services (e.g., advertisements in magazines and/or websites) via a phone call or a ghost note; the entrepreneurs are tricked into agreeing to new contracts without being aware of the obligations attached 3) A ghost invoice is a fraudulent invoice ‘Type of fraud’, in the logistic regression, was coded into two dummies: CEO fraud and Contract fraud (including both fraudulent contracts and ghost invoice) (both: 0 = no, 1 = yes) |
Primary method of communication | The contact between the fraudsters and targeted businesses is coded as 1 = email, 2 = offline post, 3 = telephone |
Routine activity concepts | |
Service orientation | Based on the type of activities as determined from the business’ website, we determined the following two variables 1) Economic sector is often operationalised in four sectors. These are: (i) primary sector—the retrieval and production of raw materials; (ii) secondary sector—the transformation of raw or intermediate materials into goods; (iii) tertiary sector—the production of services instead of end products; and (iv) quaternary or service sector—the production of knowledge-based services (Kenessey 1987) 2) Industry. Statistics Netherlands codes industrial sector according to a ‘SBI 2008—Standard Business classification’ (Statistics Netherlands 2018c), which is almost identical to its European equivalent (EUROSTAT 2008). In a logistic regression analysis, the seven largest categories (equal or higher than 5%) were recoded into dummies. There were: (i) business services; (ii) wholesale and retail trade, repairs; (iii) construction; (iv) manufacturing; (v) human health and social work activities; (vi) information and communication technology; (vii) logistics |
Business size | Business size was categorised as self-employed (coded as 1), small (coded as 2 = 2–49 employees), medium (coded as 3 = 50–99 employees), large (coded as 4 = 100–250 employees), and very large (coded as 5 = 250 employees and more) |
Business location | We coded the province in the Netherlands in which the victim was located. There are 12 provinces in the Netherlands |
Season | This was coded based on the date the fraudster first contacted the targeted company In the logistic regression ‘season of fraud attempt’ was recoded as two dummy variables ‘spring’ (0 = no, 1 = yes) and ‘summer’ (0 = no, 1 = yes), which were the two seasons that appeared to be the most relevant |
Rational choice concepts | |
Effort | 1) Crime script. The type of fraud based on its crime script. We used the ‘universal script’, from Cornish (Cornish 1993: 41, Fig. 4), leaving out the steps: ‘entry’ and ‘exit’. This includes the method of communication Primary method of communication: the contact between the fraudsters and targeted businesses is coded as 1 = email, 2 = offline post, 3 = telephone 2) Location of the bank to which the money is to be transferred. This was coded into four categories: Netherlands (coded as 1), Europe (coded as 2), Hong Kong (coded as 3), and USA (coded as 4). Originally, we coded ‘Asia’ but this appeared to be only Hong Kong, and ‘North America’, included only USA. We wanted to code the Netherlands separately, and the rest of Europe involved many different countries. For more details about the countries within Europe, see Table 6, online supplement. We assumed that the further away the bank, the more effort fraudsters put in the fraud attempt. In the logistic regression, Netherlands and Europe were combined and Hong Kong and the USA were combined into single categories |
Benefits | This is measured in terms of the amount of money requested and whether it is paid. Information on whether the business paid was also coded (0 = no, 1 = yes). The amount of money requested by the fraudsters was noted in Euro (€) In the regression analysis we used the logarithm (base 10) transformation to reduce skewness, when ‘amount of money’ is used as the dependent variable |