• No results found

4. Methodology

4.4 Measurements

The measurement scales of the variables of the research model are derived from previous literature, which was primarily written in the English language. Because the largest group of the respondents has Dutch as their native language, the measurements needed to be translated into Dutch. The back-translation procedure was used to ensure that the contents of the items remained unchanged. This means that after the items were translated into Dutch, they were translated back into English by someone else to see if the item would be the same as the original question.

The 5-point Likert scale was used for all variables with composite items. This scale was used because Qualtrics provided the insight that the dropout rate for respondents using a mobile phone when answering a 7-point Likert scale is a lot higher. This is due to the fact that the 7-point Likert scale is too large for most screens to fit and people will fatigue when answering many items using this scale. Because in this thesis the survey and experiment will mainly be distributed on social media which are used more often on the mobile phone, the 5-point Likert scale was chosen. The following section will elaborate on the used measurements.

4.4.1 Dependent variable Cross-channel free riding intention

In order to measure cross-channel free riding intentions the scale of Chou et al. (2016) was used (Cronbach’s alpha = .92). This measure consists of two items, but for the purpose of study 1, the scale was slightly adjusted. Study 1 is about previous actions, while the original scale focuses on search and purchase intentions. Therefore, the response for the item was adjusted and accepted only nominal values of either Yes or No. For study 2 however, the two items could be taken from this study. An example of the item that would be similar for both study 1 and study 2 is: “I would search through the web shop of company X, but purchase through the offline channel of another company when I buy product A.” The 5-point Likert scale was answered ranging from 1 = strongly disagree to 5 = strongly agree.

4.4.2 Independent variables Type of device

To measure the type of device, the same item was used as in the work of Kleinlercher et al. (2019).

Here they also simply asked respondents to state which device they have been using primarily to search for the products online. They provided a list of possible devices which would be: “mobile phone, tablet, laptop, desktop/PC”. Only after the data is collected, the devices have been categorized into two groups of either a fixed (laptop and desktop/PC) or a mobile device (mobile phone and tablet), like the work of De Haan et al. (2018).

33 Customer loyalty

To measure customer loyalty the scale was used like the work of Zeithaml et al. (1996) (Cronbach’s alpha = .92). Four of the five items are taken from the original work and adapted to fit in context. The measurement was the same for both studies. Two examples of items that were asked are “I say positive things about company X to other people” and “I will probably buy more products from company X in the future”. The 5-point Likert scale ranged from 1 = extremely unlikely to 5 = extremely likely.

Product type

To categorize products as either hedonic or utilitarian, the scale of Voss et al. (2003) (Cronbach’s alpha = .95) was used. The original measure consists of 12 items for both hedonic and utilitarian categories. However, only 3 were taken from each list and were selected out of the 5 highest rated based on reliability. The items in the original work lists emotions and purposes related to hedonic or utilitarian dimensions and therefore the items used in this study were adapted to fit with product categorization. The possible items for the hedonic dimension are “I buy product A because it is fun/enjoyable/exciting.” For the utilitarian dimension they are “I buy product A because it is necessary/practical/functional.” The 5-point Likert scale was answered ranging from 1 = strongly disagree to 5 = strongly agree.

Product price

In order to extract the price people have paid for their specific product, the same method was

employed as Kleinlercher et al. (2019). Here they also asked respondents for the absolute value of the price of the product they have bought in whole euros rather than giving listed categories and therefore the data that is provided is the most accurate. If the respondent did not know the exact price, they were asked to give an approximation.

4.4.3 Control variables

In the model it is important to include a control variable for the attitude people have towards online product information searching. People with a positive attitude probably have a lot of experience, and may consequently be more price conscious. This would mean that such people search more

intensively for products and discounts online. If no control variable would be added, it might be the case that the results are biased as people with a more positive attitude may use one device particularly more. Therefore, a composite variable was used that combines internet savviness and price

consciousness to reflect the attitude people have towards online shopping. Internet savviness was calculated using the scale of Maggioni et al. (2020) (Cronbach’s alpha = .76), while the scale for price consciousness was used from the work of Heitz-Spahn (2013) (Cronbach’s alpha = .89). Respondents

34

were required to answer on the 5-point Likert scale ranging from 1 = strongly disagree to 5 = strongly agree. Examples of items are: “I am good at searching for product information on the internet.” and “I often compare product prices across retailers to get the lowest possible price.”.

The other variables that will act as controls are age, gender, education, product category and the amount of time since the purchase. Age was measured using five age categories (Mathwicket al., 2001) and allows for different classifications in general life phases, which might influence their purchase behaviour. It is also important to control for education, as respondents with higher education are more likely to conduct more thorough research (Kumar & Venkatesan, 2005). The respondents answered the questions regarding the product category and how long ago the purchase was made at the beginning of the survey. The levels of education ranged from high school diploma to PhD candidate level. The gender was also extracted and could be answered in four possible ways: male, female, binary or I do not want to tell.

35