Online credit card fraud against small businesses

Kate Charlton and Natalie Taylor
ISBN 0 642 53846 8 ; ISSN 1326-6004
Canberra: Australian Institute of Criminology: 2004
(Research and public policy series, no. 60)

Appendix 1 : Methodology

Sampling process

The sampling technique involved a random sampling method, stratified by business type. Five samples (one for each of the five business types) were drawn from Australia On Disc ³ (the electronic version of the Yellow Pages telephone directories), using the appropriate ANZSIC (Australia and New Zealand Standard Industry Classification; see ABS 1993) code or category of retailer. From these five files, retailers were drawn at random until the target quotas of 200 businesses for each business type were met.⁴ The total number of completed interviews conducted was 1,078.

The survey was conducted by Roy Morgan Research, using computer-assisted telephone interviewing (CATI). Once businesses had been selected at random and contacted by telephone, they were asked two screening questions to ascertain whether they were 'out-of- scope' for the purposes of this study and should be screened out. These were (a) whether they were a retailer of the given business type, and (b) whether they employed fewer than 20 full-time personnel. The sample needed to include only retailers of the type in question, with fewer than 20 employees. However, only four per cent of businesses contacted were out of scope due to employee size, indicating that most businesses within the five business types are small businesses.

It should also be noted that since the survey was interested in online credit card fraud incidents, and it was important that the interviewee was familiar with the business' online trading practices and would know detailed information about each fraud incident, only those businesses which operated their own trading web site were classified in this study as trading online. This meant that some florists which used an external web site operated by an external company (examples include www.interflora.com.au and www.flowers.com.au) but did not operate their own separate trading web site were not classified in this report as being online (but were still interviewed). Forty-three per cent of the florists stated they accepted payments for goods sold through a web site operated by an external business. Figure A1-1 shows the original sample and outcomes to completed interviews.

Instrument

The questionnaire was administered over the telephone by a Roy Morgan trained interviewer, using the CATI technique. This was chosen as it was essential the questionnaire be interviewer-administered (the complexity and various filters and screens required computerised expertise). The telephone is both a convenient and anonymous medium for both the interviewer and interviewee.

The questionnaire incorporated a number of different sections, each of which was answered by retailers depending on whether they:

• currently traded online and had experienced fraud;
• currently traded online and had never experienced fraud;
• previously traded online; and
• never traded online.

The questionnaire is provided for reference in Appendix 2.

Piloting of survey

Prior to the survey going into the field, a pilot survey was undertaken to assess likely response rates, possible incidence rates of online trading and fraud, the utility of the questionnaire (its ability to attain useful responses), and whether retailers would have difficulty with certain questions. The pilot included 104 completed interviews (20 florists, booksellers and computer hardware retailers, 21 recorded music retailers and 23 toy and game retailers). Some minor alterations were made to the questionnaire based on the pilot phase. AIC ethics approval for the full research project was received on 18 March 2003 and the questionnaire was also approved by the Commonwealth Statistical Clearing House on 16 April 2003 after which data collection commenced.

Main data collection and response rates

Roy Morgan Research interviewers, with extensive training in social research surveys and procedures, conducted the interviews. Final data collection began on 15 May 2003 and continued until 6 June 2003. Ten interviewers worked concurrently, during as well as outside business hours (to suit the preferences of interviewees). Response rates varied little across business type, with an overall response rate for the entire sample of 62 per cent (business type response rates varied between 60% and 64%).

Sample demographics

Table A1-1 provides demographic details of the unweighted sample: the sex, size and location of businesses in each of the five types and across the entire sample. Sex varied across business type, with florists more likely to be female and computer retailers more likely to be male. Business size was largely micro, with 86 per cent of the whole sample employing fewer than five full-time staff.

Table A1-1 also provides details relating to the ARIA (accessibility/remoteness index of Australia) classification of the sample. ARIA classification is a measure of remoteness or accessibility which can be acquired from postcodes (University of Adelaide 1999) and indicates the degree to which a retailer is located in a centralised or urban area.

In this survey, remoteness was classified according to the following:

• highly accessible - an area with relatively unrestricted accessibility to a wide range of goods and services and opportunities for social interaction;
• accessible - some restrictions to accessibility of some goods, services and opportunities for social interaction; and
• remote - very restricted accessibility to goods, services and social interaction.

It is clear from the table that the vast majority (82%) of businesses were located in highly accessible areas (major towns or cities).

Weighting of data

Where data are reported in this paper which refer to the whole of Australia (for example, number of fraud incidents), the data have been weighted up to the population for greater representativeness (n=6,657). Weighted data were also used for some regression analyses (again, to ensure that the sample was representative of the distribution of the five business types across Australia), but so as not to inflate statistical significance, an effective sample size weighting was applied.

Table A1- 1: Sample demographics (unweighted data)

	Florists		Booksellers		Music retailers		Toy & game		Computer		Total
	n	%	n	%	n	%	n	%	n	%	n	%
Gender
Male	40	17.1	94	56.9	151	75.1	131	61.2	185	87.7	601	55.8
Female	194	82.9	124	43.1	50	24.9	83	38.8	26	12.3	477	44.2
(n)	(234)	(100.0)	(218)	(100.0)	(201)	(100.0)	(214)	(100.0)	(211)	(100.0)	(1,078)	(100.0)
Business size
<5 employees	226	96.6	182	83.5	174	86.6	197	92.1	149	70.6	928	86.1
5 or more employees	8	3.4	36	16.5	27	13.4	17	7.9	62	29.4	150	13.9
(n)	(234)	(100.0)	(218)	(100.0)	(201)	(100.0)	(214)	(100.0)	(211)	(100.0)	(1,078)	(100.0)
Annual turnover
<$ 200,000	125	67.6	92	49.7	62	36.0	76	40.6	49	25.4	404	43.8
$200,000 to $499,999	53	28.6	40	21.6	59	34.3	59	31.6	61	31.6	272	29.5
$500,000 to $999,999	6	3.2	32	17.3	33	19.2	33	17.6	35	18.1	139	15.1
$1 million or more	1	0.5	21	11.3	18	10.5	19	10.1	48	24.9	107	11.6
(n)	(185)	(100.0)	(185)	(100.0)	(172)	(100.0)	(187)	(100.0)	(193)	(100.0)	(922)	(100.0)
Remoteness
Highly accessible	185	80.1	185	85.3	167	83.5	159	74.6	178	85.2	874	81.7
Accessible	35	15.2	24	11.1	19	9.5	35	16.4	21	10.0	134	12.5
Remote	11	4.8	8	3.7	14	7.0	19	8.9	10	4.8	62	5.8
(n)	(231)	(100.0)	(217)	(100.0)	(200)	(100.0)	(213)	(100.0)	(209)	(100.0)	(1,070)	(100.0)
Source: Australian Institute of Criminology, Online credit card fraud against small business 2003 [computer file, unweighted data]

To weight up to the population, the data were weighted to represent the estimated total number of qualifying businesses nationally. The weighting process involved calculating the proportion of out-of-scope businesses (that is, 20 employees or more; or those not operating in the relevant industry) obtained from the sample contacted and applying this proportion to the number of businesses (post de-duplication) available from the sampling frame (Australia On Disc) for that business type.

Weighting was conducted independently within each business type, and the steps were as follows:

1. Determine the total population of each business type, as available from the sampling frame post de-duplication.
2. Subtract from that total the number of businesses known to be out-of-scope on the basis of their responses to the survey or the pilot.
3. Calculate the proportion of the attempted sample that was identified as being out-of-scope. Apply this proportion to that part of the population that was not attempted. The resulting figure is the estimated out-of-scopes among the unattempted sample.
4. Subtract the Step 3 result from the Step 2 result. The resultant figure is the estimated number of in-scope businesses within that business type.
5. The weight for each business interviewed within each business type is the Step 4 figure divided by the number of complete, in-scope interviews for that business type.

Table A1-2: Weighting steps by business type

	Florists	Book-sellers	Music retailers	Toy/game retailers	Computer retailers
Starting population (step 1)	2,932	1,228	578	755	2,077
Known out-of-scopes	42	95	76	109	206
Step 2 total	2,890	1,133	502	646	1,871
Total attempted sample	741	715	578	746	1,075
Out-of-scope as percentage of attempted	0.057	0.133	0.131	0.146	0.192
Not attempted	2,191	513	0	9	1,002
Step 3 result: Estimated out of scopes from sample not attempted	124	68	0	1	192
Step 4 result: Estimated in-scope population	2,766	1,065	502	645	1,679
Complete in-scope interviews	234	218	201	214	211
Weight (step 5)	11.82	4.88	2.50	3.01	7.96
Weight for effective sample size	1.91	0.79	0.40	0.49	1.29
Source: Roy Morgan Research, Online credit card fraud against small business: technical report (2003), unpublished

Finally, to obtain the effective sample size (for tests of statistical significance) the weights calculated in the five steps above were adjusted according to the formula, 'original weight' x (sample size/population). Table A1-2 summarises the weighting steps for each stratum.

Table A1-3 provides a breakdown of the sample by the state/territory in which they were located at the time of interviewing and by business type. The table provides both weighted and unweighted data and demonstrates the differences between each. It is clear from the table that the location of retailers with respect to state reasonably reflects the populations in each (for example, New South Wales contributed the highest number of respondents, whereas the Northern Territory contributed the fewest). With regards to differences between the weighted and unweighted data, the weighting resulted in a significant increase in florists and computer retailers and a significant decrease in music retailers and toy and game retailers - this distribution more accurately reflects the population.

Table A1-3: Respondents by business type and state - weighted and unweighted (row percentages)

State	Florists		Book-sellers		Music retailers		Toy/game retailers		Computer retailers		(n)
	UW	W	UW	W	UW	W	UW	W	UW	W	UW	W
NSW	20	39	17	14	21	9	22	11	20	27	380	2,261
Vic.	21	41	23	19	16	7	20	10	19	24	287	1,770
Qld	29	49	17	12	13	5	17	7	25	28	175	1,232
SA	23	44	20	16	22	9	20	10	16	21	101	611
WA	15	33	25	23	28	13	16	9	15	22	67	363
Tas.	24	46	22	17	20	8	17	8	17	22	41	259
ACT	22	40	39	30	13	5	9	4	17	22	23	148
NT	0	0	25	39	50	39	25	23	0	0	4	13
(n)	234	2,766	218	1,065	20	502	214	645	211	1,679	1,078	6,657
UW = unweighted, W = weighted Note: Percentages may not sum to 100 due to rounding Source: Australian Institute of Criminology, Online credit card fraud against small business 2003 [computer file, weighted and unweighted data]

---

---