Attitudinal and Situational Determinants of Self-Service Technology Use

Abstract

The authors test an extended technology acceptance model by incorporating two situational and two attitudinal variables as new predictors of self-service technology adoption. The situational variables are waiting line and service complexity, whereas the attitudinal variables are technology trust and technology anxiety. The study employs an experimental approach with hypothetical hotel check-in situations. The authors apply structural equation modeling techniques to provide additional insights into the main effects. They find significant main effects of all situational and attitudinal variables but no interaction effects between the situational and attitudinal variables. The main effects on intention to use self-service technologies occurred primarily through the mediation of perceived ease of use and perceived usefulness. The technology acceptance model needs extension to include nontechnology variables as predictors of new technology adoption.

Keywords

self-service technology technology acceptance waiting line service complexity technology trust technology anxiety

Self-service technology (SST) is becoming an increasingly inevitable part of many service operations. Today’s customers routinely encounter SST-based service options at various service stations of airlines, banks, hotels, and retail stores, among others (Kim, Christodoulidou, & Brewer, 2012). Companies deploy SSTs to either replace or complement human-based services for reasons including cost saving, market pressure, operational efficiency, and prevention of service failures (Carline, 2007; Chang & Yang, 2008; Oh, Jeong, & Baloglu, 2012; Slifka, 2010). Customers are learning to expect, desire, and choose SST-based service stations as a viable service alternative to staff-based stations to fulfill their transactional needs (Simon & Usunier, 2007). Such SSTs are infiltrating the traditionally human-focused service industry such as hotels and information centers (Cunningham, Young, & Gerlach, 2008; Riebeck, Stark, Modsching, & Kawalek, 2008; Stockdale, 2007).

A key concern arising from service innovations through SSTs relates to the conditions under which the customer desires and chooses to use an SST-based, instead of a staffed, service station. From the service provider’s standpoint, the customer’s voluntary use of SSTs for service transactions translates into return on investment, be it reduced labor costs, improved service consistency, freed staff to concentrate on other service dimensions, or satisfied customers (Erdly & Chatterjee, 2003; Meuter & Bitner, 1998). The customer, however, may or may not prefer SSTs to the service staff, depending on various personal and situational influences (Oh et al., 2012; Simon & Usunier, 2007). Consequently, one central research question has been to ask what factors affect the customer’s choice of SSTs and how the customer makes such a choice decision (e.g., Benamati, Fuller, Serva, & Baroudi, 2010; Kim & Forsythe, 2008; Lee & Rao, 2009; Lippert, 2007).

Many researchers have applied the technology acceptance model (TAM) to investigate how the customer chooses SSTs. Originally developed to understand how organizational members adopt new technologies (Davis, 1989), TAM has resulted in fruitful applications in studies of the customer’s SST choice against staff-based service delivery. For example, Kim and Forsythe (2008) studied how TAM led to actual use of sensory enabling technology in the fashion industry. Lippert (2007) examined how interorganizational and technology trusts related to TAM’s key constructs toward usage of supply chain management technology. Curran and Meuter (2005) introduced both the need-for-interaction and risk concepts as TAM’s exogenous constructs, and the former received further extensions in Oh et al.’s (2012) recent study of SSTs.

Despite the widespread use of TAM as a basic framework for understanding the customer’s SST adoption, few studies have shed light on how the newly introduced factors affected SST use through the TAM process. Most researchers have focused on the direct relationships between their newly proposed variables and the individual TAM constructs, especially SST use or use intention (e.g., Gelderman, Ghijsen, & van Dieme, 2011). Two reasons, however, necessitate analyzing TAM as a structured dependent process. First, TAM’s three key constructs (i.e., ease of use, usefulness, and intention to use) have an established causal structure among them such that a new variable’s relationship with any of TAM’s constructs should be understood in light of its relationship with the other constructs. TAM also specifies its three key constructs characterizing essentially the customer’s technology-related perceptions, and hence, the expected strong correlations among these constructs may alter the magnitude of a new variable’s impact on each construct, especially the last dependent variable in the TAM process.

This study extends previous TAM-based SST use studies in several ways. First, TAM is viewed as an SST-relevant dependent process, and data analysis illustrates how the new variables affect SST use intention through the process embedded in TAM. Second, this study introduces four exogenous variables to study the TAM process, two of which pertain to the SST adoption-specific situations and the other two to the customer’s SST-relevant attitudes. Third, this study employs an experiment to manipulate the two situational variables, while incorporating the two attitudinal variables as covariates in data analyses. Finally, this study applies structural equation modeling (SEM) techniques to analyzing the experimental data to overcome important drawbacks of such general linear models as analysis of variance and multivariate analysis of covariance (MANCOVA) that are commonly used to analyze experimental data (Bagozzi, & Yi, 1989). By doing so, this study aims to add to the SST literature, especially of hospitality and tourism, both conceptually and methodologically.

Background and Hypotheses

In this study, the independent variables are four concepts and the dependent variable is the TAM process. We first introduce and categorize the four independent variables into situational and attitudinal determinants of SST use intention through the TAM process. Situational determinants are the transactional conditions extraneous to the customer that may influence the customer’s perceptions of and intention to use an SST. They are beyond the customer’s control but routinely arise in service transaction situations. We focus on waiting line and service complexity as situational determinants. In contrast, attitudinal determinants characterize the customer’s enduring subjective knowledge, belief, perceptions, and psychological involvement related to the target SST. They often become operative when the customer thinks about using an SST for a service transaction. The two key attitudinal determinants examined in this study are the customer’s technology trust as a motivator and technology anxiety as a deterrent. Although researchers have attempted to extend TAM by introducing a variety of variables (for a critical review, see Benbasat & Barki, 2007; Lee, Kozar, & Larsen, 2003; Straub & Burton-Jones, 2007), we chose these four variables because (a) they have gained empirical support in previous studies and (b) we expected through preliminary studies that they were relevant to SST use situations at hotels. We test the effects of these four variables on SST use intention in function of the TAM process.

Technology Acceptance Model

Since its introduction (Davis, 1989; Davis, Bagozzi, & Warshaw, 1989), TAM has become a centerpiece of research models attempting to explain how employees or customers adopt new technologies. In its core, TAM aims to predict an individual’s adoption (intention) of new technologies to the extent to which the new technology is perceived to be easy to use, hence perceived ease of use, and to which it is perceived to be useful in performing a target task, hence perceived usefulness. The customer may choose to check in through an SST instead of a front desk clerk when she believes that the hotel’s SSTs are easy to use and useful in performing her pending check-in transaction (Oh et al., 2012). The customer also tends to believe SSTs are useful when they are easy to use (Benamati et al., 2010; Dimitriadis & Kyrezis, 2010; Lippert, 2007). In sum, both perceived ease of use and perceived usefulness have a direct effect on SST adoption, and ease of use additionally influences SST adoption indirectly through usefulness perceptions. This causal process of TAM is subject to the influence of a variety of extraneous factors such as transaction-specific or situational conditions and the customer’s attitudes.

Figure 1 depicts the two situational and two attitudinal variables that affect the TAM process, which is the focus of this study. Conceptualizing both attitudinal and situational variables as the antecedents of conation is consistent with attitude theories. According to the theory of reasoned action (Fishbein, 1979; Fishbein & Ajzen, 1975; Sheppard, Hartwick, & Warshaw, 1988), attitude is a key predictor of behavioral intention. Extending the theory of reasoned action, the theory of planned behavior specifies perceived control as another influential predictor of behavioral intention (Ajzen, 1991). Perceived control factors in this case may include behavioral conditions both personal, such as time and money, and situational, such as waiting line and incidental service needs. In essence, TAM is in line with the theory of reasoned action (Weijters, Rangarajan, Falk, & Schillewaert, 2007), and not surprisingly, a number of previous TAM studies are grounded in attitude theories (e.g., Benamati et al., 2010; Curran & Meuter, 2005; Lee & Rao, 2009).

Figure 1

The Conceptual Framework

TAM has also begun to provide a basic framework for related research in the hospitality discipline. Morosan (2012) introduced the concept of perceived innovativeness as an antecedent to TAM’s perceived ease of use and provided empirical evidence of its significant association with hotel guests’ intention to use biometric systems in future hotel stays. System-related features and functions added explanatory power to TAM in that hotel customers developed strong preferences for and attitudes toward computerized reservation systems when they perceived salient system attributes (Lee, Kim, & Lee, 2006). Other than TAM constructs, both extrinsic and intrinsic motivations, as well as role clarity and age, determined the customer’s intention to use self-service kiosks at hotel, casinos, and restaurants (Kim, Christodoulidou, & Brewer, 2012). TAM received a broad extension including variables such as privacy, autonomy, effectiveness, and need for interaction in explaining how hotel customers form their willingness to use SSTs instead of checking in through the front office staff (Oh et al., 2012).

Despite proliferate adoption and applications across disciplines (Lee, Kozar, & Larsen, 2003), TAM is not immune to criticism. Benbasat and Barki (2007) discuss several theoretical concerns: (a) TAM’s excessive focus on abstract constructs lacking practical implications for system design, (b) “illusion of cumulative tradition” constantly reinforcing what are already known instead of broadening new knowledge, (c) narrow focus on the predicted behavior (i.e., adoption or use of new systems), and (d) both methodological and theoretical confusion because of no commonly accepted model. Other criticisms include common method variance caused by typical within-subject study designs measuring both the independent and dependent variables concurrently and a lack of parsimony arising from a ubiquitous focus on model extensions with new variables (Straub & Burton-Jones, 2007). One desirable direction for TAM research, therefore, would be to structure influential variables into a parsimonious framework and examine thoroughly how such variables affect SST use through the TAM-defined process, and this is the focus of this study.

Situational Determinants

One frequent situational variable that may affect the customer’s SST use is waiting line (Berry, Seiders, & Grewal, 2002; Dabholkar, 1996; Dabholkar & Bagozzi, 2002). In this study, waiting line refers to the number of customers waiting to be served for their needs at the staffed service station of a hotel. The length of waiting line, which often translates into waiting time or service speed, consistently appears to be a critical issue in the hotel industry (Dabholkar, 1996; Simon & Usunier, 2007). Customers evaluate the service situation they face and strive to avoid waiting time whenever possible, thereby developing an urgent situational preference for an alternative service method such as SSTs that may minimize their waiting time (Simon & Usunier, 2007). Similarly, customers perceiving crowdedness, hence waiting, in the service area and line at airports showed a stronger intention to use SSTs (Gelderman et al., 2011). Thus, waiting time has a direct, positive effect on the customer’s intention to use alternative service solutions such as SSTs (Dabholkar, 1996). Oh et al. (2012) report a significant, positive effect of waiting line on the customer’s intent to use SSTs to reduce service congestion. We therefore predict a positive impact of waiting line on the customer’s intent to use SSTs by way of the TAM process.

Hypothesis1: Waiting line positively influences SST use intention through the TAM process.

Service complexity may also determine the customer’s intent to check in through an SST instead of a front office agent at a hotel. Service complexity is defined broadly as “the overall perception consumers have of the actual complexity of a particular service” (Simon & Usunier, 2007, p. 165). It consists of mainly two dimensions: service complexity and service divergence (Shostack, 1987). The former refers to the number and intricacy of steps required to perform the service, whereas the latter indicates the degree of freedom allowed in or inherent to a particular sequence of the service process. Service complexity was a significant, negative determinant of consumer preference for SSTs over the service personnel in government services including vehicle registration and taxation (Lee & Rao, 2009), Internet travel bookings (Anckar, 2003), and rail ticket and banking services (Anckar, Walden, & Jelassi, 2002; Simon & Usunier, 2007).

The inverse relationship between service complexity and SST usage may be tenable in hotel transactions as well. Hotel customers arrive often with a variety of service needs including, for example, room change, local travel information, and other service requests (Anckar & Walden, 2001; Cheyne, Downes, & Legg, 2006). They will strive to satisfy their service needs, and as their service needs get complicated, they may prefer the service staff to SSTs to maximize their chance for satisfaction while checking into the hotel. The customer may find the staff more flexible and versatile than SSTs and perceive SSTs increasingly inconvenient in meeting his or her service needs as the number and intricacy of steps required to perform in the pending check-in transaction grow. Thus, complicated service needs may broadly undermine SST-related perceptions and behavioral intention in favor of the service staff.

Hypothesis 2: Service complexity negatively influences SST use intention through the TAM process.

Attitudinal Determinants

In business, trust tends to engage the customer in purchase-related relationship behaviors (Johnson & Grayson, 2005; Morgan & Hunt, 1994). Trust is an attitude-like construct referring to the extent to which one has confidence in an exchange partner’s reliability and integrity (Doney & Cannon, 1997). Applied to situations of new technology adoption, technology trust is the customer’s willingness to be vulnerable to a technology based on expectations of technology predictability, reliability, and utility, and it is subject to the customer’s predilection to trust technology (Lippert, 2007; Mayer, Davis, & Schoorman, 1995; McKnight, Choudhury, & Kacmar, 2002). When the customer trusts the functionality and benefits of new technology, he or she will likely adopt and apply it for performing intended tasks (McKnight & Chervany, 2002; Pavlou & Fygenson, 2006; Turel, Yuan, & Connelly, 2008). A lack of technology trust may cause dislike and avoidance of SSTs as a method of service transactions, especially where other transaction methods such as the service staff are available. Studies have reported technology trust as a significant, attitudinal determinant of adopting a specific channel or an SST for a service transaction in Internet banking (Dimitriadis & Kyrezis, 2010), dotcom retailing (Benamati et al., 2010), and logistics operations (Lippert, 2007). Hotel guests may follow a similar processing schema when choosing to use an SST for their check-in because their trust in technology may result in broadly favorable perceptions of SSTs.

Hypothesis 3: Technology trust positively relates to SST use intention through the TAM process.

Voluntary use of new technologies demands knowledge, skills, and liking. A lack of these conditions may develop into anxiety about using the target technology, and such technology anxiety often traces back to trait, more than state, anxiety (Beckers, Wicherts, & Schmidt, 2007; Doronina, 1995). Technology anxiety arises from the inability or lack of self-confidence in effectively managing or controlling the technology and, thus, refers to the level of anxiety experienced by an individual when confronted with the decision to use a new technology (Igbaria & Parasuraman, 1989; Oyedele & Simpson, 2007). In essence, the customer lacking required technology readiness will experience the sensation of technology anxiety (Parasuraman, 2000).

Technology anxiety will discourage hotel guests to use SSTs in various service transactions including check-in. It may lead to confusion regarding the task to be performed, hence role ambiguity, decreased motivation levels, and a reduced self-perception of ability to use SSTs (Meuter & Bitner, 1998; Meuter, Bitner, Ostrom, & Brown, 2005; Ray & Minch, 1990). High levels of technology anxiety also cause low self-efficacy (Thatcher & Perrewe, 2002) and avoidance of the technology (Parasuraman, 2000; Parasuraman & Colby, 2001). Hotel lobbies where guests typically use SSTs present situations that likely augment technology anxiety because of the nature of public space. The presence of other guests may exacerbate feelings of technology anxiety for fear of potential public embarrassment, thereby deteriorating perceptions of SSTs and eventual use.

Hypothesis 4: Technology anxiety negatively relates to SST use intention through the TAM process.

Method

Study Design

We employed an experimental method based on hypothetical hotel check-in scenarios. The experiment was a 3 × 2 between-subject factorial design in which we manipulated waiting line and service complexity, while surveying technology trust and technology anxiety as covariates. Following Oh et al.’s (2012) design, the waiting line had three levels: no (0) guest versus four (4) guests versus eight (8) guests in line trying to check in at the front desk with a service representative. Service complexity had two levels, low versus high, for which we manipulated the number of required tasks to be performed during the check-in transaction (Anckar & Walden, 2001; Lee & Rao, 2009). For the low service complexity condition, the check-in process required the subject to provide a few pieces of basic information about him or her, whereas the high complexity condition framed the subject with a particular need to gather additional information about local travel options and arrange several additional hotel services. The appendix presents a sample experimental script.

We collected data electronically using a database developed recently from sampling at a regional airshow festival. The subject received an email including a cover letter and a link to the research questionnaire available on the Qualtrics™ server. We randomized assignment of the experimental conditions to the subject by using a built-in Qualtrics function. The questionnaire contained, sequentially, (a) a series of questions about the subject’s technology-related behaviors including the two research covariates, that is, technology trust and technology anxiety; (b) the experimental script, manipulation check questions, and the research questions measuring the TAM constructs; and (c) several questions about the subject’s sociodemographic information. Over an approximately 6-week data collection period, six reminders followed the original email to the nonrespondents. Incentives for study participation were two iPad 2s and 15 gift certificates (each valued at $100) through random drawings.

Sampling

We used an email database developed from the registrants to an international airshow held on August 20 and 21, 2010 in a northeastern city of the United States. The initial contact and six progressive reminders to nonrespondents resulted in 251 participants for a response rate of 22%. Of these, 11 responses were excluded from all data analyses as their responses were incomplete or contained values that appeared to be outliers. Based on 240 effective study subjects, the sample size for the experimental cells ranged from 77 (short waiting line) to 120 (low or high service complexity).

Measurement

We measured the key research variables with multiple items. Measurement of the three TAM constructs followed the multiitem approach taken by Davis (1989) and Oh et al. (2012). Two items measured the subject’s intention to choose an SST for a check-in transaction: The probability to check in through an SST was operationalized on an 11-point scale (0%, 10%, . . . , 100%) while the likelihood of choosing an SST for the check-in on a 5-point scale (1 = very unlikely, . . . , 5 = very likely). Both perceived usefulness and perceived ease of use had three measurement items each operationalized on the Likert-type scale (1 = strongly disagree, . . . , 5 = strongly agree; Oh et al., 2012). Both technology trust and technology anxiety had four items borrowed from Johnson’s (2007) and Collier and Sherrell’s (2010) studies, respectively, and they were also anchored on the same Likert-type scale (1 = strongly disagree, . . . , 5 = strongly agree).

Data Analysis

The mainstream data analysis method followed the suggestions of Bagozzi and Yi (1989) and Bagozzi, Yi, and Singh (1991), applying structural equation models to experimental studies. Bagozzi and colleagues discussed several advantages of replacing traditional general linear models such as (M)ANOVA and (M)ANCOVA with structural equation models when analyzing experimental data. For example, although the traditional methods assume homogeneity of variances and covariances of the dependent variables across the experimental groups, the structural equation modeling approach allows straightforward tests of such homogeneity assumption. Structural models also correct the variable measures for measurement errors, thereby reducing the chances of making Type II error. Another advantage of structural equation models is that they allow for a more complete modeling of underlying theoretical relationships among the variables. Structural equation models allow for step-down analyses into the set of the dependent variables having a theoretical causal order. Because our dependent variables were the three TAM constructs that had an established causal process (refer to Figure 1), application of structural equation models could overcome the drawbacks of the traditional experimental data analysis methods. For example, we could analyze the effects of the experimental variables on SST use intention while controlling for the effects of both perceived ease of use and usefulness on SST use intention. Such step-down analysis could provide useful information as to whether the mean difference in the dependent variable was due to the direct effect of the experimental manipulation or its dependency on other variables in the dependent process.

Findings

Sample and Descriptive Data

Table 1 summarizes sample characteristics. For the total sample, the majority was male (79%) and Caucasian (91%), and about three quarters reported their age to be 45 years or older. More than half the sample was holding either an associate or bachelor’s degree (53%) and earning less than $100,000 for an annual household income (57%). The sample was rather regional, with more than 86% residing in Connecticut and Massachusetts and 8% in New York. In general, the sample characteristics for the experimental subgroups (the medium and long waiting lines were combined for reasons provided below) appeared similar to those of the total sample.

Table 1

Sample Characteristics (N = 240)

			Waiting Line				Service Complexity
	Total		Short		Long		Low		High
	n ^a	%^a	n ^a	%^a	n ^a	%^a	n ^a	%^a	n ^a	%^a
Gender
Male	189	78.8	57	74.0	132	81.0	96	80.0	93	77.5
Female	49	20.4	20	26.0	29	17.8	23	19.2	26	21.7
Age
Younger than 25 years old	3	1.3	0	0	3	1.8	2	1.7	1	0.8
25-34 years old	14	5.8	4	5.2	10	6.1	10	8.3	4	3.3
35-44 years old	42	17.5	13	16.9	29	17.8	23	19.2	19	15.8
45-54 years old	65	27.1	23	29.9	42	25.8	26	21.7	39	32.5
Older than 55 years old	113	47.1	36	46.8	77	47.2	57	47.5	56	46.7
Ethnic origin
Caucasian	214	89.2	69	89.6	145	89.0	107	89.2	107	89.2
Asian American	7	2.9	3	3.9	4	2.5	1	0.8	6	5.0
Hispanic American	5	2.1	0	0	5	3.1	4	3.3	1	0.8
African American	4	1.7	1	1.3	3	1.8	3	2.5	1	0.8
American Indians	1	0.4	1	1.3	0	0	1	0.8	0	0
Education
High school or below	53	22.1	18	23.4	35	21.5	29	24.2	24	20.0
Associate degree	46	19.2	18	23.4	28	17.2	26	21.7	20	16.7
Bachelor’s degree	79	32.9	22	28.6	57	35.0	31	25.8	48	40.0
Master’s degree	47	19.6	15	19.5	32	19.6	25	20.8	22	18.3
Doctoral degree	11	4.6	4	5.2	7	4.3	6	5.0	5	4.2
Yearly household income
Less than $50,000	39	16.3	12	15.6	27	16.6	23	19.2	16	13.3
$50,000-$74,999	46	19.2	10	13.0	36	22.1	24	20.0	22	18.3
$75,000-$99,999	40	16.7	19	24.7	21	12.9	20	16.7	20	16.7
$100,000-$124,999	41	17.1	17	22.1	24	14.7	17	14.2	24	20.0
$125,000-$149,999	18	7.5	5	6.5	13	8.0	9	7.5	9	7.5
$150,000 and more	35	14.6	9	11.7	26	16.0	18	15.0	17	14.2
State of residence
Connecticut	74	30.8	22	28.6	52	31.8	36	29.9	38	31.6
Massachusetts	131	54.6	43	55.9	88	52.8	69	57.4	62	49.6
Maine	4	1.9	2	2.6	2	1.2	0	0	4	3.3
North Carolina	1	0.4	1	1.3	0	0	1	0.8	0	0
New Hampshire	6	2.4	1	1.3	5	3.0	2	1.6	4	3.3
New Jersey	2	0.8	1	1.3	1	0.6	0	0	2	1.6
New York	15	6.3	6	7.8	9	5.4	9	7.4	6	5.0
Ohio	1	0.4	0	0	1	0.6	0	0	1	0.8
Rhode Island	1	0.4	1	0.5	1	0.6	0	0	1	0.8
Vermont	5	2.1	1	1.3	4	2.4	2	1.7	3	2.4
International	2	0.8	0	0	2	1.2	0	0	2	1.6

May not sum to total due to missing responses.

Table 2 presents the results of mean comparisons at the measurement item level. Note that we collapsed the medium (4 guests; n = 82) and long (8 guests; n = 81) waiting line conditions into a “long” line as a result of a series of independent sample t tests. None of the 16 measurement items showed a significant mean difference between the two “longer” waiting line conditions, with t values ranging from −0.06 to 1.81. However, overall the short waiting line condition showed significant mean differences from both the medium and long (“longer”) line conditions, especially on the items of the three dependent TAM constructs measured immediately following the experimental treatments. Specifically, the mean comparisons between the short and medium lines resulted in t values ranging from −8.69 to −2.05, except for the technological anxiety items whose t values ranged from 0.51 to 1.84. The mean difference between the short and long line conditions was significant for all items of the TAM constructs (t = −3.09 to −8.01), three technology anxiety items (t = 2.31 to 3.42), and one technology trust item (t = −2.29). Consequently, our data analyses for the main effects hereafter focus on the condensed 2 × 2 design with the two covariates.

Table 2

Descriptive Data Results

	Waiting Line			Service Complexity
Measures^a	Short (n = 77)	Long (n = 163)	tRatio	Low (n = 120)	High (n = 120)	tRatio
Intention to choose SST for check-in
Probability to check in through SST^b	4.61 (3.04)	8.09 (2.67)	−8.59	7.55 (3.19)	6.39 (3.18)	2.82
Likelihood to choose SST for check-in^c	2.40 (1.20)	3.91 (1.08)	−9.96	3.67 (1.28)	3.18 (1.32)	2.88
Perceived usefulness
SST useful in completing check-in	3.03 (1.01)	3.96 (0.80)	−7.14	3.82 (0.99)	3.51 (0.94)	2.47
SST making my check-in smooth	3.09 (0.91)	3.64 (0.78)	−4.85	3.60 (0.83)	3.33 (0.87)	2.42
Convenient checking in through SST	3.13 (1.01)	3.98 (0.76)	−6.61	3.88 (0.94)	3.54 (0.90)	2.81
Perceived ease of use
Easy to get SST to do what I want	3.14 (0.91)	3.56 (0.85)	−3.49	3.69 (0.88)	3.17 (0.83)	4.75
Easy to use to accomplish the wanted	3.10 (0.93)	3.56 (0.88)	−3.67	3.68 (0.87)	3.14 (0.89)	4.77
Make my check-in easy and simple	3.12 (0.96)	3.63 (0.85)	−4.16	3.73 (0.89)	3.20 (0.87)	4.64
Technology trust
Trust SSTs to do the right job	3.22 (0.90)	3.44 (0.85)	−1.80	3.41 (0.87)	3.33 (0.86)	0.74
SST at hotels are reliable	3.17 (0.73)	3.43 (0.67)	−2.74	3.43 (0.71)	3.26 (0.68)	1.95
Will handle check-in without problems	3.23 (0.89)	3.50 (0.81)	−2.33	3.47 (0.81)	3.37 (0.88)	0.92
Can be trusted to handle service needs	3.09 (0.89)	3.41 (0.79)	−2.81	3.37 (0.83)	3.25 (0.84)	1.08
Technology anxiety
Hesitate to use SST, fearing mistakes	2.51 (1.03)	2.40 (0.91)	.82	2.35 (0.90)	2.52 (0.99)	−1.36
Feel insecure using new SST	2.38 (0.96)	2.03 (0.79)	2.75	2.08 (0.83)	2.20 (0.89)	−1.05
SST can be intimidating	2.29 (0.99)	1.99 (0.79)	2.33	2.08 (0.82)	2.09 (0.91)	−0.15
Use of SST can lead to embarrassments	2.25 (0.92)	2.08 (0.82)	1.36	2.13 (0.81)	2.14 (0.90)	−0.15

Note: SST = self-service technology. Entries are mean values (standard deviations).

All measures are abbreviated.

Measured on an 11-point % scale (0%, 10%, . . . , 100%).

Measured on a 5-point rating scale (1 = very unlikely, . . . , 5 = very likely); all others measured on a 5-point Likert-type scale.

Table 2 presents significant main effects as expected at the individual measure level. The means of all TAM construct items were significantly lower for the short waiting line than the long waiting line condition (p < .05); checking through an SST appears less appealing than through a service staff when the guest does not expect to wait long at the staffed service station. In contrast, all TAM items showed significantly higher means for the low than the high service complexity condition (p < .05); consequently, SSTs seem preferable when the check-in task is simple. The mean differences on the items of technology trust and anxiety vary as more than half the items resulted in insignificant t values. Note that both technology trust and anxiety were measured as covariates before the subject was treated to either of the experimental conditions. Thus, there was no particular expectation about the significance or direction of the mean differences of these constructs (or items).

Manipulation Check

Immediately after presenting the experimental scenario, we asked questions about how the subject felt about the hypothetical check-in condition. For the waiting line condition, we asked the question “The length of the waiting line at the front desk is:” on a 5-point scale with 1 = very short, 2 = short, 3 = neutral, 4 = long, and 5 = very long. The subjects assigned to the short waiting line condition, the reference group, perceived the line to be significantly shorter (M = 1.99, SD = 1.06) than those assigned to the longer line condition (M = 3.91, SD = 0.84; t = −14.0, p < .01). For a service complexity manipulation check, the question was “You feel your check-in task will be:” on a 5-point scale with 1 = very complicated, 2 = complicated, 3 = neither complicated nor simple, 4 = simple, and 5 = very simple. As expected, the low service complexity group, which was the reference group (M = 3.92, SD = 0.76), perceived the check-in task to be significantly simpler than the high service complexity group (M = 3.13, SD = 0.81; t = 7.72, p < .01). Finally, on the question “How easy or difficult is it for you to put yourself in the check-in situation described above?” the mean value was relatively high (M = 3.95, SD = 0.91) on the 5-point scale with 1 = very difficult, 2 = difficult, 3 = neutral, 4 = easy, and 5 = very easy. Thus, the manipulation plan of this study seemed to be realistic as intended.

Measurement Model

Using the total sample, we evaluated the convergent and discriminant validities of the key model constructs to assure that the multiitem measures provided a sound measurement framework for hypothesis tests. The measurement model fit the data well with χ² = 153.36 (df = 116), p = .012, root mean square error of approximation (RMSEA) = .04, comparative fit index (CFI) = .99, and nonnormed fit index (NNFI) = .99. As shown in Table 3, the factor loadings of each construct are generally high and significant, whereas the error estimates are relatively small, providing evidence for convergent validity of the constructs (Anderson & Gerbing, 1988; Bagozzi & Yi, 1988). Reliability of the constructs is also evident in both the ρ_η and Cronbach’s α values that are higher than .70 (Bagozzi & Yi, 1988; Fornell & Larcker, 1981; Nunnally, 1978). The amount of variance extracted for each construct (ρ_vc(η)) exceeded the suggested minimum of .5 (Bagozzi & Yi, 1988; Fornell & Larcker, 1981). We assessed discriminant validity of the model constructs by comparing the correlation of each pair of the constructs to the square root of the ρ_vc(η) value for each of the two constructs in the correlation (Fornell & Larcker, 1981). The correlation coefficients were smaller than the square root of the involved ρ_vc(η) values in all cases, which substantiated discriminant validity of the constructs. Combined with our additional examination of all individual parameters to assure a proper solution of the measurement model, these results provide necessary evidence for both convergent and discriminant validities of the model constructs.

Table 3

The Measurement Model Results

	Factor Loading		Error
Latent and Observed Variables	Estimate^c	t Ratio	Estimate^c	t Ratio	ρ_η^a	ρ_vc(η)^b	α
Intention to choose SST for check-in
Probability to check in through SST	0.94	—	0.12	4.96	0.94	0.88	0.94
Likelihood to choose SST for check-in	0.94	24.91	0.11	4.67
Perceived usefulness
SST useful in completing check-in	0.90	—	0.20	7.73	0.91	0.76	0.91
SST making my check-in smooth	0.85	18.39	0.28	8.94
Convenient checking in through SST	0.88	19.81	0.23	8.25
Perceived ease of use
Easy to get SST to do what I want	0.95	—	0.09	6.72	0.96	0.89	0.96
Easy to use to accomplish the wanted	0.94	30.57	0.11	7.56
Make my check-in easy and simple	0.93	29.61	0.13	7.98
Technology trust
Trust SSTs to do the right job	0.87	—	0.25	7.97	0.92	0.73	0.91
SST at hotels are reliable	0.84	16.62	0.30	8.68
Will handle check-in without problems	0.83	16.28	0.31	8.85
Can be trusted to handle service needs	0.88	18.22	0.22	7.47
Technology anxiety
Hesitate to use SST, fearing mistakes	0.80	—	0.36	9.18	0.91	0.72	0.91
Feel insecure using new SST	0.86	15.20	0.26	8.04
SST can be intimidating	0.91	16.16	0.18	6.43
Use of SST can lead to embarrassments	0.82	14.23	0.33	8.91

Note: SST = self-service technology. Both waiting line (coded as 0 and 1) and service complexity (0, 1) were included in the model, but they are not presented here because their loadings were fixed at 1.0 and errors at 0.

a, b. Construct reliabilities and amount of variance extracted, respectively, based on Fornell and Larcker’s (1981) formula.

Standardized values.

Tests of Research Hypotheses

The effects of waiting line on the TAM constructs including SST use intention appear to be significant as summarized in Table 4. In the table, column One’s indicates a pseudo variable having a column of constant value 1s in the raw data to correctly compute the likelihood function and standard errors of estimates when analyzing the means of observed dependent variables as a function of the categorical independent variables (Sörbom, 1974; Bagozzi & Yi, 1989). The categorical or experimental variable, waiting line, was dummy coded 0 (short) and 1 (longer), and hence, the parameters from it to the TAM constructs indicated the mean difference in the respective TAM constructs. We analyzed the augmented moment matrix rather than the usual correlation or covariance matrices. The one-way MANCOVA-equivalent structural equation model with both technology trust and technology anxiety as covariates fit the data acceptably (χ² = 342.02, df = 116, p = .0, RMSEA = .09, CFI = .97, and NNFI = .96; Hu & Bentler, 1999). Note that this model specified the dependent TAM constructs as correlated rather than causally ordered in order to show through a step-down analysis later how a causal order among the dependent variables could change the effects of the independent variables (i.e., waiting line in this case). Following the logic of χ²-difference analyses (Bagozzi & Yi, 1989; Bagozzi et al., 1991), we tested the significance of the effects of waiting line on the three TAM constructs by constraining the three parameters simultaneously to zero. This constrained model resulted in a significantly worse model (Δχ² = 92.72, df = 3, p < .001) than the unconstrained baseline model, indicating that waiting line had a significant effect on the TAM (constructs) collectively. The mean values of the TAM measures were consistently higher in the longer waiting line condition than those in the short line. Ad hoc traditional MANCOVA analysis also confirmed a significant waiting line effect (i.e., mean difference) on the TAM constructs simultaneously (Wilks’s Λ = .64, p < . 001), in the expected direction.

Table 4

One-Way MANCOVA Results for Waiting Line, With Correlated TAM Constructs

	Independent Variables
Dependent Variables	Waiting Line	One’s^a	Technology Trust	Technology Anxiety	Line × Trust	Line × Anxiety
Ease of Use	.06^b	.93	.11	−.05	−.01	−.02
	2.41^c	39.93	7.00	−3.14	−.39	−1.41
Usefulness	.14	.87	.08	−.05	−.03	−.02
	6.50	40.51	5.60	−3.34	−2.36	−1.49
SST Use Intent	.29	.70	.09	−.08	.01	−.02
	9.13	21.25	4.23	−3.58	.48	−.69
χ²	434.74^d		393.59^d	358.68^d	649.65^e	638.42^e
Δχ^2f	92.72*		51.57*	16.66*	14.18*	2.95^ns

Note: MANCOVA = multivariate analysis of covariance; TAM = technology acceptance model; SST = self-service technology; RMSEA = root mean square error of approximation; NNFI = nonnormed fit index; CFI = comparative fit index.

A pseudo variable of constant 1s in the raw data to model group means correctly.

Standardized estimate.

t Ratio.

The baseline model without the interaction terms: χ² = 342.02, df = 116, p = .0; RMSEA = .09; NNFI = .96; CFI = .97.

The baseline model with the interaction terms: χ² = 635.47, df = 265, p = .0; RMSEA = .08; NNFI = .96; CFI = .97.

df = 3.

p < .01. ns = not significant (p > .05).

Disaggregate analyses of the waiting line effects on the TAM constructs also affirmed Hypothesis 1. The mean difference in ease of use between the short and longer waiting lines was significant and positive (β = .06, t = 2.41, p < .05; see Table 4). A significant, positive mean difference between the two line conditions was evident as well for perceived usefulness (β = .14, t = 6.50, p < .01) and intention to use SSTs (β = .29, t = 9.13, p < .01). Ad hoc traditional MANCOVA analyses supported these results in that the longer waiting line had a significantly larger mean than the short waiting line on ease of use (Wilks’s Λ = .93, p < .001), perceived usefulness (Wilks’s Λ = .77, p < .001), and SST use intention (Wilks’s Λ = .71, p < .001). Therefore, Hypothesis 1 could not be rejected.

We assessed and found support for the effect of technology trust, as a covariate, on the TAM constructs (Hypothesis 3). When imposed with the assumption that the effects of technology trust on the three TAM constructs simultaneously be null, the fit of the one-way structural MANCOVA model deteriorated significantly compared with that of the baseline model (Δχ² = 51.57, df = 3, p < .001; see Table 4). According to the individual parameter results in Table 4, technology trust positively related to ease of use (β = .11, t = 7.00, p < .001), perceived usefulness (β = .08, t = 5.60, p < .01), and intention to use SSTs (β = .09, t = 4.23, p < .01), respectively. Hypothesis 3 could not be rejected.

Table 4 also summarizes the results supporting a significant effect of technology anxiety on the TAM constructs. An alternative model with the three parameters of technology anxiety effects on the TAM constructs resulted in a significantly inferior model fit than that of the baseline model (Δχ² = 16.66, df = 3, p < .01). Individually, the technology anxiety effect was significant and negative, as expected, on ease of use (β = −.05, t = −3.14, p < .01), perceived usefulness (β = −.05, t = −3.34, p < .01), and intention to use SSTs (β = −.08, t = −3.58, p < .01). These results provided evidence to support Hypothesis 4.

The interaction effects of (a) waiting line by technology trust and (b) waiting line by technology anxiety on the TAM constructs were generally insignificant. We extended analyses into the potential interaction effects, albeit not hypothesized because of the absence of theoretical support, to explore additional determinants of the TAM constructs. In essence, these interactions could reflect how attitudinal variables, technology trust, and technology anxiety reacted to situational variables, waiting line. As shown in Table 4, the model including these two interaction variables resulted in an acceptable model fit: χ² = 635.47 (df = 265), p = .0, RMSEA = .08, CFI = .97, and NNFI = .96. An alternative model, with the three interaction parameters of waiting line × technology trust constrained to zero, entailed a significantly worse model fit (Δχ² = 14.18, df = 3, p < .01). Such a significant decrease in model fit was due mainly to the significant, negative interaction effect of waiting line × technology trust on perceived usefulness (β = −.03, t = −2.36, p < .05); the interaction effects on ease of use (β = −.01, t = −.39, p > .05) and SST use intention (β = .01, t = .48, p > .05) were not significant. The model with the three waiting line × technology anxiety effects constrained to zero did not significantly undermine the model fit (Δχ² = 2.95, df = 3, p > .05). Individually, the interaction variable had no significant effect on the three TAM constructs (p > .05; see the last column of Table 4). Based on these results, we did not consider interaction effects in further analyses.

Parallel analyses provided support for a significant effect of service complexity on the TAM constructs. As in Table 5, the one-way MANCOVA model produced an acceptable model fit: χ² = 339.52 (df = 116), p = .0, RMSEA = .09, CFI = .96, and NNFI = .95. The model constraining the three effects of service complexity on the TAM constructs to zero produced a significantly worse fit to the data (Δχ² = 26.35, df = 3, p < .01). Ad hoc traditional MANCOVA analysis supported this result (Wilks’s Λ = .88, p < .001), with the mean values of TAM measures higher for the low service complexity condition than for the high service complexity condition. Each parameter was also significant (p < .01) and negative as expected (see the first entry column in Table 5). TAM construct-specific traditional MANCOVA analyses, too, produced consistent results revealing higher mean values for the low service complexity condition in ease of use (Wilks’s Λ = .90, p < .001), perceived usefulness (Wilks’s Λ = .96, p ≈ .05), and intention to use SSTs (Wilks’s Λ = .96, p ≈ .03). Combined, these results provide empirical support for Hypothesis 2.

Table 5

One-Way MANCOVA Results for Service Complexity, With Correlated TAM Constructs

	Independent Variables
Dependent Variables	Service Complexity	One’s^a	Technology Trust	Technology Anxiety	Complexity × Trust	Complexity × Anxiety
Ease of Use	−.09^b	1.00	.11	−.05	.03	.02
	−5.08^c	58.05	7.24	−3.44	1.54	1.21
Usefulness	−.04	1.00	.09	−.06	−.01	−.01
	−2.46	54.12	5.73	−3.62	−1.23	−.77
SST Use Intent	−.08	.99	.11	−.10	−.03	.00
	−2.62	32.25	4.38	−3.38	−2.00	.24
χ²	365.87^d		389.93^d	360.16^d	610.21^e	603.49^e
Δχ^2f	26.35**		50.41**	20.64**	8.12*	1.40^ns

A pseudo variable of constant 1s in the raw data.

Standardized estimate.

t ratio.

The baseline model without the interaction terms: χ² = 339.52, df = 116, p = .0; RMSEA = .09; NNFI = .95; CFI = .96.

The baseline model with the interaction terms: χ² = 602.09, df = 265, p = .0; RMSEA = .07; NNFI = .94; CFI = .95.

df = 3.

p < .01. *p < .05. ns = not significant (p > .05).

Table 5 provides additional evidence for the effects of technology trust and technology anxiety on the TAM constructs. The model constraining the three technology trust effects on the TAM constructs to zero deteriorated the model fit significantly (Δχ² = 50.41, df = 3, p < .001). All three individual parameters were statistically significant and positive (p < .001). For technology anxiety, the model with the three parameters constrained to zero resulted in a significantly worse fit (Δχ² = 20.64, df = 3, p < .001). Each parameter was significant and negative (p < .001). These results are consistent with the waiting line–based results above, adding to the empirical evidence for Hypotheses 3 and 4.

The interaction effects based on the service complexity model generally appeared to be in line with those based on the waiting line model earlier. The model including the two two-way interaction variables (i.e., complexity × trust and complexity × anxiety) fit the data acceptably (χ² = 602.09, df = 265, p = .0, RMSEA = .07, CFI = .95, and NNFI = .94; see Table 5). A model constraining the three complexity × trust effects to zero resulted in a marginally worse fit (Δχ² = 8.12, df = 3, p ≈ .04); only the interaction effect of task complexity × trust on SST use intention was marginally significant (β = −.03, t = −2.0, p ≈ .04), whereas the other two were not significant (p > .05). For the complexity × anxiety interaction effect, a model constraining the three parameters to zero did not undermine the model fit (Δχ² = 1.40, df = 3, p > .10); none of the three parameters was individually significant (p > .05). Overall, these interaction effect results supported the elimination of such effects in further analyses.

Step-Down Analyses

Step-down analyses provide useful information because they test whether variation in a certain dependent variable is due to a direct association with the manipulation or due to its dependence on other dependent variables in the model (Bagozzi & Yi, 1989). Finding mean differences for different groups is one thing, but uncovering how the manipulation effect interacts with an a priori ordering of the dependent variables provides an even more powerful and insightful set of information. Given the well-established causal ordering of the TAM constructs (see Figure 1), step-down analyses could provide additional useful insights, beyond what we reported above, into how the two experimental and covariate variables affected SST use intention through the TAM process. To this end, we fit a one-way MANCOVA model to the data by using a multigroup structural equation analysis approach. Not only does the multigroup approach provide cross-group or group-specific information, but it also overcomes the implicit assumption of homogeneity between the experimental groups in traditional MANCOVA analyses.

Table 6 summarizes the results of the waiting line effect when the TAM constructs are causally ordered. The two-group, one-way MANCOVA model fit the data acceptably (χ² = 397.45, df = 210, p = .0, RMSEA = .09, CFI = .96, and NNFI = .94). A model of measurement equivalence (i.e., invariant factor loadings; see Byrne, 2008) between the two waiting line groups did not significantly undermine the model fit (Δχ² = 19.2, df = 11, p > .05). We tested the mean difference in the TAM constructs between the two waiting line groups by comparing the measurement equivalence model with a model imposing an equality assumption on the same parameters between the two groups. This invariant means model worsened the model fit significantly (Δχ² = 19.81, df = 3, p < .01), indicating the means of the TAM constructs were different between the two groups. Examinations of individual parameters revealed that when the TAM constructs were modeled as ordered, waiting line affected ease of use and usefulness perceptions only (p < .01), but not SST use intention (β = −.05, t = −.23, p > .05, for the short line; β = −.24, t = −1.85, p > .05, for the longer line). These results substantiated that waiting line affected SST use intention only through perceived ease of use and perceived usefulness, following the TAM-embedded process. They provided empirical support for the part “through the TAM process” in Hypothesis 1.

Table 6

One-Way Multigroup MANOCOVA Model Results for Waiting Line, With Causally Ordered TAM Constructs^a

	Waiting Line		Trust		Anxiety		Usefulness		Ease of Use
TAM Constructs	Short	Long	Short	Long	Short	Long	Short	Long	Short	Long
Ease of Use	.99	.96	.15	.09	−.03	−.07
	34.7	64.15	4.43	4.54	−.86	−3.43
Usefulness	.26	.37	.04	−.01	−.02	−.02			.74	.64
	3.17	6.59	1.63	−1.11	−.79	−1.27			9.24	11.15
SST Use Intent	−.05	−.24	−.03	.05	−.06	−.01	.79	1.50	.20	−.30
	−.23	−1.85	−.58	1.98	−1.23	−.06	2.12	5.90	.60	−1.52

The model fit is as follows: χ² = 397.45, df = 210, p = .0; RMSEA = .087; NNFI = .94; CFI = .96.

The order effect was even more dramatic with technology trust and technology anxiety variables. The effect of technology trust was significant only on ease of use under both the short (β = .15, t = 4.43, p < .01) and longer (β = .09, t = 4.54, p < .01) line conditions, whereas the effect on perceived usefulness and SST use intention was insignificant (p > .05). Thus, the trust effect on both usefulness and SST use intention could be because of their dependence on ease of use. The effect of technology anxiety on the TAM constructs was not significant under the short waiting line condition (p > .05), whereas it was significant and negative only on ease of use under the longer line condition (β = −.07, t = −3.43, p < .01). A significant, positive effect was present from ease of use to usefulness and from usefulness to SST use intention but not from ease of use to SST use intention directly, under both waiting line conditions. These results lend additional evidence supporting the process aspects stated in Hypotheses 3 and 4.

Table 7 presents the results of the one-way, two-group MANCOVA model for service complexity when the TAM constructs are ordered. The model fit the data acceptably (χ² = 322.74, df = 210, p = .0, RMSEA = .07, CFI = .95, and NNFI = .94). The measurement equivalence model (i.e., equal factor loadings between the groups) did not make the model fit worse (Δχ² = 1.11, df = 11, p ≈ 1.0). A model constraining the means to be equal between the two groups was not acceptable (Δχ² = 63.17, df = 3, p < .001). Disaggregate observations indicate that the mean difference was significant for both ease of use and usefulness but not for SST use intention. Under both low and high service complexity conditions, technology trust significantly affected ease of use only, but not usefulness and SST use intention. Similarly, the effect of technology anxiety was significant and negative only on ease of use, but not on usefulness and SST use intention when the TAM constructs were ordered. The direct effect of ease of use on SST use intention was not significant, whereas the other relationships among the TAM constructs were significant and positive. These findings collectively supported the part “through the TAM process” in Hypotheses 2, 3, and 4.

Table 7

One-Way Multigroup MANCOVA Model Results for Service Complexity, With Causally Ordered TAM Constructs^a

	Service Complexity		Trust		Anxiety		Usefulness		Ease of Use
TAM Constructs	Low	High	Low	High	Low	High	Low	High	Low	High
Ease of Use	.99	.97	.09	.13	−.06	−.05
	46.62	61.21	3.22	6.87	−2.30	−2.93
Usefulness	.26	.23	−.02	.02	−.03	−.01			.74	.76
	3.98	2.76	−.89	1.01	−1.76	−.10			11.17	8.90
SST Use Intent	−.09	−.10	.07	−.01	.02	−.04	1.83	1.29	−.48	−.22
	−.47	−.72	1.82	−.45	.60	−1.67	5.13	6.07	−1.60	−1.00

The model fit is as follows: χ²=322.74, df = 210, p = 0.0; RMSEA = .067; NNFI = .93; CFI = .95.

Discussion and Implications

The key findings of this study are the significant situational and attitudinal effects on the customer’s choice of SSTs as a transaction method at hotels. Both waiting line and task complexity as situational variables appeared strong determinants of SST use intention for checking in at hotels. As the waiting line at the staffed front desk gets longer, the customer tends to favor SSTs over the service staff for the check-in transaction (Hypothesis 1). In contrast, the customer coming with more service demands tended to avoid using SSTs (Hypothesis 2). Thus, the customer seems to favor SSTs for speedy, less elaborate transactions. SSTs may not appeal when there are only few people waiting to be served by the service staff and when the customer needs to fulfill a number of information needs at check-in. Hotel operators trying to introduce SSTs should expect their return on investment with careful consideration given to these situational or operational conditions. SST designers also need to focus on the desired utility of SSTs as a speedy and simple transaction device, rather than trying to build excessive, but unused capacities in SSTs. Travelers, such as business travelers, who travel frequently and are used to a variety of SSTs, may find these systems quick, easy, and speedy in completing their transactions and getting to their room faster.

The customer’s trust in and anxiety about technology are attitudinal variables determining the customer’s choice of SSTs for check-in transactions. The stronger the customer trusts technology, the more likely he/she will intend to use SSTs (Hypothesis 3). Conversely, the anxiety the customer has about technology tends to discourage him/her to use SSTs (Hypothesis 4). Such effects of technology trust and anxiety arose consistently in both the direction and strength when they were modeled together with waiting line (Table 4) and service complexity conditions (Table 5). These findings support those of previous studies in other service settings (Johnson, Bardhi, & Dunn, 2008; Meuter, Ostrom, Bitner, & Roundtree, 2003) and add pieces of evidence for the same effects occurring in hotel transactions. Operators deploying SSTs may increase SST usage rates, hence return on investment, by helping the customer appreciate and embrace the advantages of using SSTs and, at the same time, reduce the customer’s apprehension about using SSTs. Relevant consumer education via various channels as well as operational tactics to boost consumer trust and alleviate consumer anxiety about SSTs is in order. SST aids for first-time users may help reduce the anxiety and speed up the service cycle as well.

Dynamic relationship structures present in the dependent variables require careful analyses and interpretation of the effects of the independent variables, both experimental and correlational. Experimental studies relying on the traditional general linear models usually examine main effects on often more than one dependent variable while ignoring the theoretical relationships among the dependent variables. Step-down analyses in this study revealed that the causal order among the dependent variables as in the three TAM constructs could provide different interpretations of the main effects. That is, when the effect of a dependent variable(s) on another dependent variable(s) is controlled, the main effect on this latter dependent variable could either diminish or disappear. In this study, the effects of the two situational and attitudinal variables on intention to use SSTs showed the patterns that justified implementation of step-down analysis. Either ease of use or the combination of ease of use and usefulness could mediate completely the effects of both the experimental (i.e., situational) and correlational (i.e., attitudinal) variables on SST adoption. Thus, in the presence of an established theoretical structure among the dependent variables, reliance on the traditional analysis methods may lead to a narrow and often erroneous result in understanding of the main effects. This is particularly true because in practice the dependent variables used in the same experimental study often derive from the same theory or they at least correlate significantly with each other. Practitioners need to correctly understand whether the independent variables predict the dependent variables directly or indirectly for effective resource allocation decisions. SST use intention relies on the SST’s ease of use, which in turn could be a function of the trust and anxiety the customer has toward to the SST.

Methodologically, step-down analysis could be a practical alternative to formal mediation analysis. For example, both step-down and mediation analyses attempt to uncover how the intervening variables affect the direct relationship between the independent and dependent variables. Both analyses require clearly defined causal relationships among the variables, whereas step-down analysis additionally requires a clearly defined dependent structure or process. Traditional mediation analysis models the intervening or mediating variables, such as perceived ease of use and usefulness, as the independent variables in analysis steps (Baron and Kenny, 1986), whereas step-down analysis models them as the dependent variables (Bagozzi & Yi, 1989; Bagozzi et al., 1991). Given the reliance on the prespecified theoretical dependent process, step-down analysis may be easier to justify for application because the intervening variables are already known and their relationship with the key dependent variables are defined by theory. In addition, it is relatively easier (two steps) to implement than mediation analysis (three steps).

Although widely adopted, TAM needs greater extensions and further analysis in future application studies. The TAM constructs are bound to a specific target technology under consideration of deployment or adoption. They are descriptive of the target technology without considering an array of other situational and attitudinal variables that may determine the goal of TAM, that is, predicting use of the focal technology (Davis, 1989). Different technologies and their specific adoption situations may introduce different sets of key determinants that may not necessarily relate to the characteristics of the target technology. Using an SST use situation in hotel check-in transactions as an example, this study showed how some situational and customer attitude variables could affect SST use intention. Researchers may often need exploratory studies to cultivate relevant determinants in different SST use situations. Practitioners need to be careful of interpreting past study results and consider their specific decision-making situations to correctly understand influential variables specific to the SST use situation in hand.

In addition to uncovering situational and attitudinal determinants, future studies need to improve TAM on prescriptive dimensions. TAM’s constructs are highly abstract and conceptual, which makes them suitable for examining the underlying theoretical relationships. A focus on such abstract concepts, however, offers limited opportunities to generate practical ideas that can assist system engineers in designing easily adoptable SSTs. One promising avenue may be to develop the antecedents of perceived ease of use and usefulness that characterize the features and functions of the target SST system such that study results can entail specific prescriptions for how to design easy-to-use and useful SSTs (Benbasat & Barki, 2007). Alongside this practical direction, commensurate efforts are necessary to advance TAM in a high level of model parsimony (Straub & Burton-Jones, 2007). Parsimony has been a lost concept in a large number of previous studies attempting to extend TAM. For managers and system designers alike, model parsimony as well as explanatory power is an important parameter in the decision economics. This study attempted to show in a way how to structure extension variables into two parsimonious categories, that is, situational and attitudinal. A next desirable task would be to determine the most influential variables within each category and model them together with key system feature variables in an effort to improve the practical utility of TAM.

More multistage and longitudinal studies are necessary to advance TAM (Benbasat & Barki, 2007). Most previous TAM-based studies relied on one-shot surveys or experiments and measured all constructs concurrently (Lee et al., 2003). Such concurrent measurement of the independent and dependent variables could run a risk of common method variance and memory spillover (Straub & Burton-Jones, 2007). The common method bias becomes more serious when the multiple constructs share the same measurement scale. Operationalizing constructs simply on different scales may reduce method variance artifacts significantly. Measuring or observing actual SST use separately in time from measuring attitudinal antecedents such as perceived ease of use and usefulness may prevent method variance problems as well, albeit more challenging in study design and research costs.

Finally, based on the set of the variables used, this study detected no interaction between the situational and attitudinal variables toward intention to use SSTs. The length of waiting line did not interact with either the customer’s trust in or anxiety about technology to significantly encourage or discourage the customer to favor SSTs over the service staff for a hotel check-in transaction. Service complexity, too, did not significantly augment or discount the customer’s intention to use SSTs by interacting with either the customer’s technology trust or anxiety. Theoretical efforts need to continue to uncover not only spanning boundaries for the main effects of the two situational and attitudinal variables but also other key situational and attitudinal determinants of SST use and their potential interactive effects. Knowledge about interaction effects will help make operational strategies more effective.

Conclusion

This study provides empirical evidence for the roles played by the four variables, especially in the lodging industry. We conceptualized these variables into situational or attitudinal to provide a more parsimonious category structure to the key determinants of SST use and thereby to offer a general conceptual framework for future research. The main effects of the two situational and two attitudinal variables were evident, but their interaction effects did not evince. We also demonstrated application of structural equation modeling for analyzing experimental data containing a theoretically ordered dependent process. Such step-down analyses provided insights into how the newly introduced independent variables influenced SST use intention through the TAM process, the results that have not been available through previous experimental studies relying on traditional analysis methods. We further argue that TAM provide only a narrow window of the new SST adoption process and hence need greater efforts to include powerful system-related and nontechnology predictors. In that sense, this study provided an organized starting point for structuring variables and analyzing data.

Future research may build on limitations of this study. First, the study scope may extend including different SSTs and other SST usage situations both within the hotel industry and across different industries. Today, technology infiltrates service operations across the board at an increasing rate, and thus, understanding key situational and attitudinal variables associated with different SST use contexts would be valuable. In addition, other situational and attitudinal variables such as presence of other customers in line to use SSTs (Oh et al., 2012) and social anxiety (Dabholkar & Bagozzi, 2002) deserve attention in future studies.

Second, future studies need to collect data from more representative samples if external validity of the study results is a primary goal (Lynch, 1982). This study’s sample was skewed largely toward an older, male, Caucasian, northeast residence, and special travel purpose demographic. A younger, business, or frequent traveler sample may yield different results. Although random samples drawn from a wider sampling frame could have been more desirable, such sampling may not always be desirable for studies emphasizing theory tests (Calder, Phillips, & Tybout, 1982). We focused on collecting data from a traveling consumer group with potential needs for accommodation where they might run into opportunities to use SSTs. Future studies may pursue the same design goal with broader sample bases to improve both generalizability and theory testing.

Finally, actual users of SSTs versus staff-based services may be surveyed or interviewed to understand motivations and conditions under which they chose a particular method of transaction. An interview or focus group study of recent users of both SSTs and staff services may offer a good opportunity to compare the reasons and motivations for choosing one against the other transaction method. In addition, when some interaction effects are present and when the sample size of each experimental cell is large enough to warrant valid results, application of structural equation modeling to experimental data may extend to multigroup analyses to test various interaction hypotheses. Such efforts will advance theoretical progress in this increasingly growing area of research and industry practice.

Footnotes

Appendix

Haemoon Oh (e-mail: oh@isenberg.umass.edu) is a professor and department head, Department of Hospitality and Tourism Management, Isenberg School of Management, University of Massachusetts–Amherst. Miyoung Jeong (e-mail: mjeong@isenberg.umass.edu) is an associate professor in the Department of Hospitality and Tourism Management, Isenberg School of Management, University of Massachusetts–Amherst. Seonjeong (Ally) Lee (e-mail: seonjeon@som.umass.edu) is a PhD student in the Department of Hospitality and Tourism Management, Isenberg School of Management, University of Massachusetts–Amherst. Rodney Warnick (e-mail: warnick@isenberg.umass.edu) is a professor in the Department of Hospitality and Tourism Management, Isenberg School of Management, University of Massachusetts–Amherst.

References

Anckar

. (2003). Consumer intentions in terms of electronic travel distribution implications for future market structures. e-Service Journal, 2(2), 68-86.

Anckar

Walden

. (2001). Self-booking of high- and low-complexity travel products: Exploratory findings. Information Technology & Tourism, 4, 151-165.

Anckar

Walden

Jelassi

. (2002). Creating customer value in online grocery shopping. International Journal of Retail and Distribution Management, 30, 211-220.

Anderson

Gerbing

. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103, 411-423.

Ajzen

. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-211.

Bagozzi

. (1988). On the evaluation of structural models. Journal of the Academy of Marketing Science, 16(1), 74-94.

Bagozzi

. (1989). On the use of structural equation models in experimental designs. Journal of Marketing Research, 26, 271-284.

Bagozzi

Singh

. (1991). On the use of structural equation models in experimental designs: Two extensions. International Journal of Research in Marketing, 8, 125-140.

Beckers

Wicherts

Schmidt

. (2007). Computer anxiety: “Trait” or “state”? Computers in Human Behavior, 23, 2851-2862.

10.

Benamati

Fuller

Serva

Baroudi

. (2010). Clarifying the integrating of trust and TAM in e-commerce environments: Implications for systems design and management. IEEE Transactions on Engineering Management, 57, 380-393.

11.

Benbasat

Barki

. (2007). Quo vadis, TAM? Journal of the Association for Information Systems, 8, 211-218.

12.

Berry

Seiders

Grewal

. (2002). Understanding service convenience. Journal of Marketing, 66(3), 1-17.

13.

Byrne

B. M

. (2008). Testing for multigroup equivalence of a measuring instrument: A walk through the process. Psicothema, 20, 872-882.

14.

Calder

Phillips

Tybout

. (1982). The concept of external validity. Journal of Consumer Research, 9, 240-244.

15.

Carline

. (2007). Competitive advantage: Consumers are more ready for kiosks than operators. In Executive Summary of Hospitality Industry Self-Service Technology Study 2007. Retrieved from http://c3185012.r12.cf0.rackcdn.com/v3-docs/4e1881ef61ec03f72d234e5b415c5429/rhss_htstudy07_ar.pdf

16.

Chang

Yang

. (2008). Do airline self-service check-in kiosks meet the needs of passengers? Tourism Management, 29, 980-993.

17.

Cheyne

Downes

Legg

. (2006). Travel agent vs. internet: What influences travel consumer choices? Journal of Vacation Marketing, 12, 41-57.

18.

Collier

Sherrell

. (2010). Examining the influence of control and convenience in a self-service setting. Journal of the Academic Marketing Science, 38, 490-509.

19.

Cunningham

Young

Gerlach

. (2008). Consumer view of self-service technologies. Service Industries Journal, 28, 719-732.

20.

Curran

Meuter

. (2005). Self-service technology adoption: Comparing three technologies. Journal of Services Marketing, 19, 103-113.

21.

Dabholkar

. (1996). Consumer evaluations of new technology-based self-service options: An investigation of alternative models of service quality. International Journal of Research in Marketing, 13(1), 29-51.

22.

Dabholkar

Bagozzi

. (2002). An attitudinal model of technology-based self-service: Moderating effects of consumer traits and situational factors. Journal of Academy of Marketing Science, 30, 184-201.

23.

Davis

. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13, 318-340.

24.

Davis

Bagozzi

Warshaw

. (1989). User acceptance of computer technology: A comparison of two theoretical models. Management Science, 35, 982-1003.

25.

Dimitriadis

Kyrezis

. (2010). Linking trust to use intention for technology-enabled bank channels: The role of trusting intentions. Psychology & Marketing, 27, 799-820.

26.

Doney

Cannon

. (1997). An examination of the nature of trust in buyer-seller relationships. Journal of Marketing, 61(2), 35-51.

27.

Doronina

. (1995). Fear of computers: Its nature, prevention, and cure. Russian Social Science Review, 36(4), 76-95.

28.

Erdly

Chatterjee

. (2003). The case for self service in hospitality. Retrieved from http://www.hotel-online.com/News/PR2003_4th/Oct03_SelfServiceTech.html

29.

Fishbein

. (1979). A theory of reasoned action: Some applications and implications. Nebraska Symposium on Motivation, 27, 65-116.

30.

Fishbein

Ajzen

. (1975). Belief, attitude, intention, and behavior: An introduction to theory and research. Reading, MA: Addison-Welsey.

31.

Fornell

Larcker

D. F

. (1981). Evaluating structural equation models with unobservable variables and measurement error. Journal of Marketing Research, 18(1), 39-50.

32.

Gelderman

Ghijsen

van Diemen

. (2011). Choosing self-service technologies or interpersonal services: The impact of situational factors and technology-related attitudes. Journal of Retailing and Consumer Services, 18, 414-421.

33.

Bentler

. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.

34.

Igbaria

Parasuraman

. (1989). A path analytic study of individual characteristics, computer anxiety and attitudes toward microcomputers. Journal of Management, 15, 373-388.

35.

Johnson

. (2007). Achieving customer value from electronic channels through identity commitment, calculative commitment, and trust in technology. Journal of Interactive Marketing, 21, 2-22.

36.

Johnson

Grayson

. (2005). Cognitive and affective trust in service relationships. Journal of Business Research, 58, 500-507.

37.

Johnson

Bardhi

Dunn

. (2008). Understanding how technology paradoxes affect customer satisfaction with self-service technology: The role of performance ambiguity and trust in technology. Psychology & Marketing, 25, 416-443.

38.

Kim

Christodoulidou

Brewer

. (2012). Impact of individual differences and consumers’ readiness on likelihood of using self-service technologies at hospitality settings. Journal of Hospitality & Tourism Research, 36, 85-114.

39.

Kim

Forsythe

. (2008). Adoption of virtual try-on technology for online apparel shopping. Journal of Interactive Marketing, 22(2), 45-59.

40.

Lee

Kim

Lee

. (2006). Testing the determinants of computerized reservation system users’ intention to use via a structural equation model. Journal of Hospitality & Tourism Research, 30, 246-266.

41.

Lee

Kozar

K. A.

Larsen

K. R. T

. (2003). The technology acceptance model: Past, present, and future. Communications of the AIS, 12, 752-780.

42.

Lee

Rao

. (2009). Task complexity and different decision criteria for online service acceptance: A comparison of two e-government compliance service domains. Decision Support Systems, 47, 424-435.

43.

Lippert

. (2007). Investigating postadoption utilization: An examination into the role of interorganizational and technology trust. IEEE Transactions on Engineering Management, 54, 468-483.

44.

Lynch

. (1982). On the external validity of experiments in consumer research. Journal of Consumer Research, 9, 225-239.

45.

Mayer

Davis

Schoorman

. (1995). An integrative model of organizational trust. Academy of Management Review, 20, 709-734.

46.

McKnight

Chervany

N. L

. (2002). What trust means in e-commerce customer relationships: An interdisciplinary conceptual typology. International Journal of Electronic Commerce, 6(2), 35-59.

47.

McKnight

Choudhury

Kacmar

. (2002). Developing and validating trust measures for e-commerce: An integrative typology. Information Systems Research, 13, 334-359.

48.

Meuter

Bitner

. (1998). Self-service technologies: extending service frameworks and identifying issues for research. In Grewal

Pechmann

(Eds.), Marketing theory and applications (pp. 12-19). Chicago, IL: American Marketing Association.

49.

Meuter

Ostrom

Bitner

Roundtree

. (2003). The influence of technology anxiety on consumer use and experience with self-service technologies. Journal of Business Research, 56, 899-906.

50.

Meuter

Bitner

Ostrom

Brown

. (2005). Choosing among alternative service delivery modes: An investigation of customer trial of self-service technologies. Journal of Marketing, 69, 61-83.

51.

Morgan

Hunt

. (1994). The commitment-trust theory of relationship marketing. Journal of Marketing, 58(3), 20-38.

52.

Morosan

. (2012). Theoretical and empirical considerations of guests’ perceptions of biometric systems in hotels: Extending the technology acceptance model. Journal of Hospitality & Tourism Research, 36, 52-84.

53.

Nunnally

. (1978). Psychometric theory. New York, NY: McGraw-Hill.

54.

Jeong

Baloglu

. (2012). Tourists’ adoption of self-service technologies at resort hotels. Journal of Business Research, 66, 692-699.

55.

Oyedele

Simpson

. (2007). An empirical investigation of consumer control factors on intention to use selected self-service technologies. International Journal of Service Industry Management, 18, 287-306.

56.

Parasuraman

. (2000). Technology readiness index: A multiple-item scale to measure readiness to embrace new technologies. Journal of Service Research, 2, 307-320.

57.

Parasuraman

Colby

. (2001). Techno: Ready marketing, how and why your customers adopt technology. New York, NY: Free Press.

58.

Pavlou

Fygenson

. (2006). Understanding and predicting electronic commerce adoption: An extension of the theory of planned behavior. Management Information Systems Quarterly, 30, 115-144.

59.

Ray

Minch

. (1990). Computer anxiety and alienation: Toward a definitive and parsimonious measure. Human Factors: The Journal of the Human Factors and Ergonomics Society, 32, 477-491.

60.

Riebeck

Stark

Modsching

Kawalek

. (2008). Studying the user acceptance of a mobile information system for tourists in the field. Information Technology & Tourism, 10, 189-199.

61.

Sheppard

Hartwick

Warshaw

. (1988). The theory of reasoned action: A meta-analysis of past research with recommendations for modifications and future research. Journal of Consumer Research, 15, 325-343.

62.

Shostack

. (1987). Service positioning through structural change. Journal of Marketing, 51, 34-43.

63.

Simon

Usunier

. (2007). Cognitive, demographic, and situational determinants of service customer preference for personnel-in-contact over self-service technology. International Journal of Research in Marketing, 24, 163-173.

64.

Slifka

. (2010). Self-service in the hospitality industry. Retrieved from http://www.selfservice.org/article_4848_25.php

65.

Sörbom

. (1974). A general method for studying differences in factor means and factor structure between groups. British Journal of Mathematical and Statistical Psychology, 27, 229-239.

66.

Stockdale

. (2007). Managing customer relationships in the self-service environment of e-tourism. Journal of Vacation Marketing, 13, 205-219.

67.

Thatcher

Perrewe

. (2002). An empirical examination of individual traits as antecedents to computer anxiety and computer self-efficacy. Management Information Systems Quarterly, 26, 381-396.

68.

Turel

Yuan

Connelly

. (2008). In justice we trust: Predicting user acceptance of e-customer services. Journal of Management Information Systems, 24, 123-151.

69.

Straub

D. W.

Jr Burton-Jones

. (2007). Veni, Vidi, Vici: Breaking the TAM logjam. Journal of the Association for Information System, 8, 223-229.

70.

Weijters

Rangarajan

Falk

Schillewaert

. (2007). Determinants and outcomes of customers’ use of self-service technology in a retail setting. Journal of Service Research, 10(1), 3-21.