Chapter 4: Research methodology
4.6 Error control process
As identified by Visser et al. (2000) and Zikmund (2010), several types of errors can
occur in the survey processes. These errors can be divided into four types: sampling
errors or coverage errors, respondent errors and measurement errors. The sampling or
coverage errors occur when the sample does not properly represent the target population
This can occur due to economic reasons, such as not considering the whole population
from which to select the sample, or the results obtained from the survey can be biased to
one section of the sample (Hair et al. 2011; Zikmund 2010). Sampling errors or
coverage errors were minimised in the stage one survey, as the whole population was
included in the survey. To establish the correct contact details of the respondents, the
latest version of the directory was used. Therefore, all the information regarding the
respondents was up-to-date. The in-person interview sample was selected using the data
received from the stage one survey. Further, when selecting the prospective
respondents, their work experience, job purview and their authority to represent the
organisation were also considered. These steps helped to minimise sampling or
coverage errors.
Non-response errors are very common in mail surveys, however these can also occur in
in-person interviews due to several reasons, for example, the prospective respondent
refuses to participate in the survey, or the appropriate respondent is not selected for the
survey, or due to physical factors, such as the respondent is unable to respond to the
mail questionnaire (Hair et al. 2011; Zikmund 2010). In order to minimise non-response
errors with the mail survey and prior to contacting prospective respondents for in-
explained the objectives and the importance of the respondents‟ contribution to the surveys, which enhanced the response rate by reducing anxiety. To reduce the risk of
mail being lost, all the mail and the advance letters were posted from Sri Lanka.
Furthermore, each package included a postage-paid self-addressed envelope in order to
improve the response rate. The option of email was also offered so the respondents
could return their responses electronically if they so wished. This minimised non-
response errors, as the respondents were made to feel that they were given an easy
option to respond.
As a strategy to minimise non-response errors in in-person interviews, each respondent
was offered the findings of the stage one survey . Furthermore, during the confirmatory
telephone conversation, each respondent‟s agreement to participate in the interview was sought and if they required any clarification regarding the interview this was provided
prior to the interview. To deter any early termination, verbal and non-verbal
encouragements, as explained in the administration process were used. According to
Franke and Franke (1977), individual response bias errors occur when the respondent
knowingly provides incorrect information. This error was minimised by enlisting the
most suitable respondents for the survey (Krosnick 1999). During the in-person
interview at the beginning of Section G, respondents were informed that the interview
was close to completion to retain respondents‟ attention. Duffy et al. (2005) explain that in-person interview respondents are more susceptible to social bias due to the presence
of an interviewer. To minimise the social bias in the responses, every respondent was
informed that there was no right or wrong answer and only their perceptions were
considered (Duffy et al. 2005). Furthermore, respondents were informed by the
respondents. This was to make them more comfortable in explaining their experience
(Leigh, Claude and Martin 1987).
The non-response error can occur due to measurement errors in the mail survey and the
in-person interview questionnaire (Visser, Krosnick and Lavrakas 2000). Measurement
errors can occur due to formatting and administration issues in questionnaires (Phillips
1981). To minimise formatting errors in questionnaires, steps were taken to reduce
ambiguity and enhance the clarity of the questions (De Vaus 2002). As mentioned under
questionnaire development (section 4.4.3), several types of unsuitable questions were
not included in the questionnaires. All the instructions given in the questionnaires were
kept clear and uniformly simple. Furthermore, to minimise memory error during the in-
person interview, response cards were given to respondents (Oishi 2003). The questions
were properly pronounced by the interviewer during the in-person interview and several
probing techniques, as mentioned in the in-person interview administering section
(section 4.4.8), were used to enhance the quality of answers and to avoid early
termination. To ensure interviewer alertness during the interview, only two interviews
were scheduled per day to minimise any error due to interviewer fatigue.
In the mail survey, non-response bias was analysed with the aim of assessing the
generalisability of the sample population (Armstrong & Overton, 1977). Ideally, the
strongest test of non-response bias was to contact non-respondents and compare the data
from them. However, this option could not be performed due to financial constraints. As
a result, non-response bias was dealt with by comparing early respondents with late
respondents. As proposed by Armstrong and Overton (1977), this method is based on an
respondents. That is, some prospective respondents are too busy, or uneasy about filling
in a survey. Accordingly, comparison between early and late respondents reveals that
there are no significant differences on key demographic variables (Appendix O-Item A).
4.7
Summary
This chapter explained the reasons for the selection of the two stage data gathering . The
mail survey was selected as it was more suitable for an exploratory study and the in-
person interview was selected to provide an in-depth component to data gathering . For
the stage one mail survey, the total population (84 senior managers) representing global
container lines and their agencies were selected. Based on the responses received from
them, 36 senior managers were selected for the stage two in-person interviews
representing four types of strategic co-operations. The detailed process of developing
questionnaires to suit the selected data gathering methods and the processes adopted to
pre-test these questionnaires were also discussed in this chapter. In addition, this chapter
has discussed the questionnaire administration and error control processes adopted by
the researcher to enhance the data quality. The next chapter discusses the findings of the
stage one mail survey followed by an analysis of findings of the interviews in Chapter
Six.
CHAPTER FIVE
RESOURCES IN CONTAINER LINES
Chapter 5: Resources in container lines