Cd

Journal of Business Research 64 (2011) 292–298

Contents lists available at ScienceDirect

Journal of Business Research

Comparing theories to explain e-commerce adoption

Elizabeth E. Grandón a,⁎, Suzanne A. Nasco b, Peter P. Mykytyn Jr. b

a College of Business Administration, Universidad del Bío-Bío, Av. Collao 1202, Concepción, Chile b College of Business Administration, Southern Illinois University, Mailcode 4629, Carbondale, IL 62901-4629, United States

⁎ Corresponding author. E-mail addresses: [email protected] (E.E. Grandó

(S.A. Nasco), [email protected] (P.P. Mykytyn).

0148-2963/$ – see front matter © 2009 Elsevier Inc. Al doi:10.1016/j.jbusres.2009.11.015

a b s t r a c t

a r t i c l e i n f o

Article history: Received 1 March 2009 Received in revised form 1 September 2009 Accepted 1 November 2009 Available online 24 December 2009

Keywords: E-commerce Theory comparison Structural equation models Chile

E-commerce is a strategy for rapid growth, especially by small and medium sized businesses (SMEs). However, the adoption rate of e-commerce by SMEs in Latin America is still undersized. The authors compare the theory of planned behavior (TPB) and the theory of reasoned action (TRA) using structural equation modeling to determine which is better at predicting e-commerce adoption intentions among 210 SME managers/owners in Chile. Contrary to previous research with American respondents, the study does not find significant differences between the two theories. Thus, academics should select the more parsimonious model (TRA) to study e-commerce adoption issues in developing countries.

n), [email protected]

l rights reserved.

© 2009 Elsevier Inc. All rights reserved.

1. Introduction

The theory of planned behavior (TPB) (Ajzen, 1991) is an extension of the theory of reasoned action (TRA) (Fishbein and Ajzen, 1975; Ajzen and Fishbein, 1980). Both theories hypothesize that an individual's intention to perform the behavior in question is a determinant of that behavior. Intentions are “indications of how hard people are willing to try, of how much of an effort they are planning to exert, in order to perform the behavior” (Ajzen, 1991, p. 181). Thus, the individual's attitude toward the behavior and the subjective norm determine intention. Attitude toward the behavior refers to the degree to which a person has a favorable or unfavorable evaluation of the behavior in question. Subjective norm refers to the perceived social pressure to perform or not to perform the behavior.

In 1991, Ajzen (1991) added a new construct to the TPB: perceived behavioral control (PBC). He states that the original model (TRA) was unable to deal with behaviors over which people have incomplete volitional control. PBC reflects an individual's perceptions that personal and situational impediments to the performance of the behavior may exist. Ajzen (1991) argues that the more favorable the attitude and subjective norm with respect to a behavior, and the greater the perceived behavioral control, the stronger should be an individual's intention to perform the behavior under consideration.

Numerous researchers (c.f., Sheppard et al., 1988; Madden et al., 1992; Randall, 1994; Orbell et al., 1997) use the TPB and TRA to predict behavioral intentions and/or behavior in several academic disciplines. Many studies find that the TPB has a better explanatory

power than its counterpart, the TRA, due to the addition of the perceived behavioral control construct (Madden et al., 1992; Taylor and Todd, 1995a; Venkatesh et al., 2003). However, a lack of research results in uncertainty when considering whether these findings apply to other types of behavioral intentions and contexts. Thus, the purpose of this study is to corroborate the predictive validity of these two theories in a different research setting — the intention to adopt e- commerce by managers/owners of SMEs in Chile.

2. Theoretical background

Many researchers publishing in MIS journals (e.g., Lee et al., 2006; Pavlou and Fygenson, 2006; Fu et al., 2006; Pee et al., 2008; Khalifa and Shen, 2008; Nor and Pearson, 2008) use the TPB, and by default the TRA, to explain behavioral intention to adopt information technology. However, only a small portion of these studies focus on samples extracted from SMEs, such as Harrison et al. (1997); Riemenschneider and McKinney (2001–2002); Riemenschneider et al. (2003) and Nasco et al. (2008). Harrison et al. (1997) use the TPB to predict small business executives' decisions to adopt information technology to achieve a competitive advantage. Normal- ly, researchers follow a three-step process: completion of an elicitation study for the preliminary TPB questionnaire, identification of computer-based systems firms in the future will adopt, and customization of the final questionnaire sent out to executives in SMEs. Results often find strong support for the TPB theory based on attitude, subjective norms, and perceived behavioral control regard- ing IT adoption.

Also testing the TPB, Riemenschneider and McKinney (2001– 2002) analyze the beliefs of small business executives regarding the adoption of e-commerce. They find that all of the component items of

293E.E. Grandón et al. / Journal of Business Research 64 (2011) 292–298

the normative and control beliefs differentiated between adopters and non-adopters of e-commerce. In the behavioral beliefs (attitude) group, however, only some items (e.g., e-commerce enhances the distribution of information, improves information accessibility, communication, and the speed with which things get done) are found to differentiate adopters from non-adopters of e-commerce.

Following a similar line of inquiry, Riemenschneider et al. (2003) examine the factors that influence web site adoption by SMEs. To gain an understanding of how small businesses handle IT adoption issues, they propose a combined model using the TPB and the technology acceptance model (TAM) of Davis (1989). They test individual models, partially integrated models, and fully integrated models through structural equation modeling techniques and find that the fully combined model provides a better fit than either the TPB or the TAM alone. However, the perceived behavioral control construct of the TPB and the perceived usefulness construct of the TAM are not significant predictors of web adoption in the full model.

Recently, Nasco et al. (2008) use the TPB to model intentions to adopt e-commerce among managers/owners of SMEs in Chile. This novel study conducted in a developing country finds support for the theory. Using hierarchical regression, the authors show that the subjective norm and attitude constructs positively and significantly predict the intention to adopt e-commerce. Contrary to hypotheses, perceived behavioral control does not significantly influence man- agers/owners´ intention to adopt the technology in question. As that study only uses regression and does not focus on a true theory comparison, stricter tests are necessary. Table 1 summarizes these studies.

3. TPB vs. TRA

Madden et al. (1992) test the predictive validity of the TRA and the TPB for ten different behaviors (exercising regularly, getting a good night's sleep, talking to a close friend, doing laundry, avoiding caffeine, going shopping with a friend, renting a videocassette, taking vitamin supplements, listening to an album, and washing one's car), utilizing responses from 94 business students. Results indicate that inclusion of PBC enhances the prediction of behavioral intention and

Table 1 E-commerce-related MIS adoption research in SMEs using the TPB.

Source Critical influencing factors

Adoption measure No. of SMEsa

Partici

Harrison et al. (1997) Attitudes Adopt new IT 162 SMEs (nb200)

Busine Subjective norms

Perceived behavioral controls

Riemenschneider and McKinney (2001–2002)

Attitudeb E-commerce adoption

184 SMEs (nb500)

Presid and mSubjective norms

Perceived behavioral controls

Riemenschneider et al. (2003)

Attitudes Web site adoption (web presence)

156 SMEs (nb500)

CEOs, generaSubjective norms

Perceived behavioral controlns

Perceived usefulnessns

Nasco et al. (2008) Attitudes E-commerce adoption

212 SMEs (nb200)

Manag Subjective norms

Perceived behavioral controlns

s: significant variable (pb0.1); ns: non-significant variable (p≥0.1). a n represents the maximum number of employees considered in the criteria to define a b Partial significance (not all the items in the construct were significant).

actual behavior. Ajzen (1991) summarizes 16 studies and concludes that the addition of PBC to the model leads to considerable improvements in the prediction of intention.

Few studies in the MIS field directly compare the TPB to the TRA to empirically test model fit as these theories apply to the adoption of IT. Chang (1998), Taylor and Todd (1995a), and Venkatesh et al. (2003) are some exceptions. Chang (1998) examines whether 181 university students in Hong Kong would make unauthorized software copies, whereas Taylor and Todd (1995a) examine whether business school students in a midsize university in the USA would use the computer resource center. Both studies find similar results: the TPB, with the PBC construct, has a better model fit than the TRA.

Venkatesh et al. (2003) also finds support for the PBC construct. The authors compare the TPB and the TRA, among other models, to determine their powers in predicting behavioral intentions among individuals in the workplace to adopt information systems, such as an online meeting manager, database applications, portfolio analyzers, and proprietary accounting systems. Four organizations participated over a six-month period, with three points of measurement (post-training, one month after implementation, and three month after implementation). The TPB explains 37%, 25%, and 21% respectively, at each time point, while the TRA explains slightly lower variance amounts (30%, 26%, and 19% respectively). Thus, the inclusion of the PBC construct appears to add to the prediction of behavioral intentions in the MIS environment.

However, none of these studies compares the TPB and the TRA in international contexts Thus, researchers should determine whether the research in American settings regarding the ability to predict behavioral intentions related to e-commerce adoption applies to different cultures of the world. Moreover, no study examines the predictive ability of both theories in a developing country. Thus, building on earlier work (Nasco et al., 2008), this project compares the TPB and TRA in predicting e-commerce adoption by owners and managers of small businesses in a developing country, Chile.

3.1. Cultural differences

Hofstede's (1997) research on cultural dimensions provides a theoretical underpinning that may help to explain differences in e-

pants Industry

ss executives Service, non-profit, and manufacturing

ent/CIO, vice president, managers, IS arketing directors, and others

Defense, agriculture, oil and gas, and manufacturing

COOs, vice presidents, IS managers, and l managers

Service/sales, government, retail, banking, medical, and manufacturing

ers and owners Wholesale, retail, construction, manufacturing, transportation, and others

SME.

Fig. 1. Model 1 represents the TRA (full lines). Model 2 represents the standard TPB (full lines plus dotted lines).

294 E.E. Grandón et al. / Journal of Business Research 64 (2011) 292–298

commerce adoption between developed and developing countries. Surveying employees from the IBM Multinational Corporation in 72 countries, including the USA and Chile, Hofstede identifies four dimensions to distinguish among different cultures: power distance, individualism, masculinity, and uncertainty avoidance.

The USA scores substantially higher than Chile on individualism (91 vs. 23) and masculinity (63 vs. 28). Higher individualism, defined as “the interest of the individual prevails over the interest of the group” (Hofstede, 1997, p. 50), indicates that a culture is more interested in individual goals and pursuits, rather than communal needs of the larger social group. Hofstede defines masculinity as pertaining to “societies in which social gender roles are clearly distinct” (p. 82); higher values of masculinity suggest that traditional gender roles and dominant societal values such as assertiveness, acquisition of money, and focus on material success are present in such a society.

Chile, on the other hand, scores higher than the USA on power distance (63 vs. 40) and uncertainty avoidance (86 vs. 46). Hofstede defines power distance as “the extent to which the less powerful members of institutions and organizations within a country expect and accept that power is distributed unequally” (Hofstede, 1997, p. 28). In a working environment, larger values of power distance mean consider- able dependence of employees on bosses, where employees are unlikely to approach and contradict their bosses directly. Finally, larger values on the uncertainty avoidance dimension, defined as “the extent to which the members of a culture feel threatened by uncertain or unknown situations” (p. 113), reflect cultures whose members who behave in ways that avoid or prevent feelings of uncertainty.

Differences in Hofstede's cultural dimensions between Chile and the USA may influence, to a certain extent, e-commerce adoption intentions. In this regard, uncertainty avoidance and individualism may have the most direct bearing on such perceptions. According to Hofstede's classification, the US culture may be more prone to risk taking and willingness to assume changes, while Chile, on the other hand, exhibits a culture that is less prone to risk taking and, in general terms, may seek to avoid change. The incorporation of a new technology, such as e-commerce, brings structural changes and redesign of organizations (Laudon and Laudon, 2004), which Chilean individuals may not be as comfortable with as their American counterparts. In addition, the American culture is very individualistic, which may dictate the willingness of US managers/owners of SMEs toward adopting an electronic business model, such as e-commerce. More collectivist cultures, such as Chile, may perceive e-commerce as an impersonal way to conduct business, antithetical to the way a collectivistic culture perceives business relationships.

Many researchers apply the TPB and the TRA to American firms, with the theories demonstrating good explanatory power (Taylor and Todd, 1995a; Harrison et al., 1997; Riemenschneider et al., 2003). Thus, determining the extent to which these theories apply to a collectivist culture with high uncertainty avoidance index and low individualism index is still an issue.

3.2. Research model and determinants of the theories

The TPB (Ajzen, 1991) states that behavior (B) is a direct positive function of behavioral intention (BI) and perceived behavioral control (PBC). PBC influences behavior indirectly through intentions, as well as directly when the person does not have complete control over that behavior and when the individual's perceptions of control are accurate (Madden et al., 1992). Fig. 1 graphically shows the direct effect of PBC on behavior and its indirect effect through intentions. Behavior is a weighted function of intention and PBC (B=w1BI+ w2PBC). Mykytyn and Harrison (1993) indicate that “assuming nothing has arisen in the environment to cause a change in plans, a measure of intention should be the best predictor of behavior” (p. 17). One's attitude (A), subjective norm (SN), and perceived behavioral

control (PBC) determine behavioral intention (BI). Thus, according to the TPB, BI is a weighted function of A, SN, and PBC, i.e., BI=w1A+ w2SN+w3PBC.

This study tests two competing models representing the ante- cedents of intention. Fig. 1 shows Models 1 and 2: Model 1 (full lines) represents the TRA where the only determinants of intention are attitude and subjective norm. Model 2 (full and dotted lines) incorporates PBC as the third factor influencing intention. Model 2 nests Model 1 when the path from PBC to intention is equal to 0.

4. Research methodology

4.1. Subjects

Top managers/owners in this study are from a population of SMEs in Chile. The researchers identified a random sample of 1100 small businesses chosen from various business directories in the capital city of Santiago, the Bío-Bío, and the Ninth region of Chile. The Chilean Corporation to Promote Production defines a small business as one that employs between 10 and 200 employees (CORFO, 1994). The sample firms receive a printed survey, which includes a cover letter, the questionnaire, and a pre-paid return envelope. Follow-up telephone calls to non-respondents are made two, four, and six weeks after the surveys were mailed. Thirty-five firms did not receive the survey instruments due to incorrect addresses. Over a 12-week period, the researchers receive 228 surveys, representing a response rate of 20.27%.

The TPB and TRA reflect with individual behaviors, but the theories are applicable to SMEs by considering that the respondents are top managers/owners who are the primary decision makers in their respective firms. Similar to Harrison et al.'s (1997) work, in this study, the researchers carefully screened respondents to ensure that each person is a primary decision-maker in his/her firm. The researchers eliminate any response from a firm from the final analysis if the firm does not meet the definition of SMEs, if the respondent is not a top manager, or if the firm had e-commerce already in place. As a result, the final analysis contains 212 firms.

4.2. Measures

Measuring latent constructs of attitude, subjective norm, percep- tions of behavioral control, and intention in the context of the TPB and TRA involves direct measures that represent manifest indicators. The items in the present instrument include appropriately modified items from the study by Riemenschneider et al. (2003) to reflect the specific target behavior, the intention to adopt e-commerce by managers/ owners of SME in Chile. To measure the TPB, five items measure attitude, four items measure subjective norm, three items measure perceived behavioral control, and three items measure intention. Measurement of all the items is on a 7-point Likert scale, ranging from

295E.E. Grandón et al. / Journal of Business Research 64 (2011) 292–298

(1) strongly agree to (7) strongly disagree. Contact the first author for the original Spanish questionnaire or the English version of the items.

5. Results

5.1. Measurement model

The researchers conducted confirmatory factor analyses (four separate CFAs involving one for each construct) and Cronbach's alpha reliability analyses prior to testing the competing TPB and TRA models to examine the internal consistency of the items representing each construct. For the hypothesized models with three indicators (intention and perceived behavioral control), the models are saturated (the number of parameters is equal to the number of variances and covariances among the observed variables) (Byrne,

Table 2 Results of CFA for intention, attitude, subjective norm, and PBC.

Construct and indicators Completely standardized loadings

Construct reliability (α) and indicator reliability

Variance extracted

Intention 0.96a 0.88a

I1. My firm strongly intends to incorporate e-commerce within the next year

0.90 0.81 0.19

I2. We have certain plans to incorporate e-commerce in our organization within the next year

0.98 0.96 0.04

I3. Our firm has a strong commitment to incorporating e-commerce within the next year

0.94 0.88 0.12

Attitude 0.88a 0.71a

A1. Incorporating e-commerce in my firm within the next year would be good

0.87 0.76 0.24

A3. Incorporating e-commerce in my firm within the next year would be positive

0.91 0.83 0.17

A4. Incorporating e-commerce in my firm within the next year would be effective

0.73 0.53 0.47

Subjective norm 0.90a 0.70a

SN1. Most people who are important to my firm think my firm should incorporate e-commerce within the next year

0.94 0.88 0.12

SN2. Most people who influence the behavior of my firm think my firm should incorporate e-commerce within the next year

0.86 0.73 0.27

SN3. People whose opinions our firm value would prefer our firm to incorporate e-commerce within the next year

0.87 0.75 0.25

SN4. Most firms that are important to my firm have adopted e-commerce

0.67 0.45 0.55

Perceived behavioral control 0.70a 0.57a

PBC1. Incorporating e-commerce in my firm within the next year would be easy

0.87 0.76 0.24

PBC2. Incorporating e-commerce in my firm within the next year would be under my firm's control

0.55 0.30 0.70

PBC3. Incorporating e-commerce within the next year would be simple to arrange

0.81 0.66 0.34

a Construct reliability and variance extracted at the construct level respectively.

1998). Thus, the model fit is perfect (χ2 (0)=0.00, pb1.00). More important than fit, the t-values associated with each indicator are statistically significant. The researchers measured the subjective norm latent construct with four observed indicators. The CFA for this construct produces satisfactory results with χ2 (2)=2.79, (pb0.25) and Root Mean Square Error of Approximation (RMSEA)=0.04. Other model fit indices also indicate quite strong values for a well-fitting measurement model (GFI=0.99, AGFI=0.97; RMSEA=0.04).

The results of the CFA for the attitude construct are initially not acceptable. Two indicators (the second and fifth) from the measure- ment of this construct have high error variances. Coincidentally, both of these items are reverse-worded, such that high numbers meant lower attitudes and the meaning of reverse-worded items may be difficult for international respondents to understand. Thus, the model for attitude now contains only 3 indicators (A1, A3, and A4). As three indicators lead to a saturated solution, the re-specified model has a perfect fit (χ2 (0)=0.00, pb1.00) and the t-values associated with each item are statistically significant.

Table 2 shows completely standardized loadings, reliability and variance extracted for each indicator. Bold numbers indicate construct reliability and variance extracted at the construct level.

5.2. Structural models

The research team used structural equation modeling (SEM) to test the validity of the competing models in this study by examining the relationships between the theoretical constructs. Statistical model comparisons provide contrasting model fit indices and allow for statistical tests between models, such as likelihood ratios (χ2

difference test) and Wald tests (Bollen, 1998).

5.2.1. The full model (TPB) The first structural model examined is the full TPB model, with

three exogenous constructs and one endogenous construct. The model includes 13 observed indicators (10 for the exogenous constructs and 3 for the endogenous construct). Fig. 2 presents the tested model with LISREL notations, showing path coefficients between each construct and its respective indicators (λ) and the error terms associated with each indicator (δ and ε). Gamma coefficients (γ) represent the paths from the exogenous constructs to the endogenous construct, which correspond to regression coefficients among the constructs. Contact the first author for the covariance matrix used as SEM input.

Within the initial SEM program, the Phi (PH) matrix is a symmetric matrix with all parameters freely estimated and the Theta Delta (TD) and Theta Epsilon (TE) matrices are diagonal with all parameters

Fig. 2. Initial theoretical structural model (SEM notation).

Fig. 4. The TRA model with path indicators (Model 1).

296 E.E. Grandón et al. / Journal of Business Research 64 (2011) 292–298

freely estimated. Exogenous variables correlate, which is standard practice in SEM (Loehlin, 1998; Byrne, 1998). The Gamma (GA) matrix is a full matrix with all parameters fixed to zero. Subsequently, gamma components (1,1), (1,2), and (1,3) are free to be estimated. Lambda-X (LX) and Lambda-Y (LY) matrices are set as full matrices with all parameters fixed to zero. The researchers used the first indicator of each latent variable to set the unit of measurement.

Output statistical goodness-of-fit measures show that the initial model is not well-fitting: the chi-square value is significant, χ2 (59)= 198.81 pb0.001, and other indices are poor (RMSEA= 0.11, GFI=0.87, AGFI=0.79, and χ2/df=3.37). Inspecting the error vari- ance matrix shows that the second indicator PBC2 has high error variance and also has the lowest R2 value of 0.32. Therefore, the researchers dropped this item from the analysis and re-estimate the model with 12 indicators.

After re-estimation, the results show a slight decrease in χ2 (196 with 48 degrees of freedom). The RMSEA, however, results in a higher value (0.12) when compared with the RMSEA of the previous model with 13 indicators. Other goodness-of-fit indices result in poorer fit (GFI=0.86, and AGFI=0.78, χ2/df=4.09). A closer look at indicator A4 shows that the variable has a high error variance of 1.07, which contributes to the misfit of the model. Therefore, a follow-up analysis did not include indicator A4.

Finally, the researchers respecify the model with 11 indicators. This model shows a better fit to the data; the chi-square value drops considerably to 106.91 with 38 degrees of freedom (pb0.001). Although this value is significant, the χ2/df ratio is very good at 2.81. All of the other goodness-of-fit statistics also support the revised model (RMSEA = 0.094, GFI = 0.91, AGFI = 0.85, RMR = 0.03, CFI=0.99, NNFI=0.98, NFI=0.98). LISREL output indicates that attitude, subjective norm, and perceived behavioral control in the revised structural model for the TPB explain 81% of the variance in intention. Fig. 3 depicts the final revised model along with the standardized paths.

5.2.2. The reduced model (TRA) The TRA does not include the PBC construct; thus this model is a

reduced model. According to Bentler (1992), nested models emerge when constraints exist on a more general model, creating a reduced model. Thus, the TRA model is a nested model within the TPB when the direct path between the latent exogenous construct of PBC to the endogenous construct of intention is set to zero. To compare the models, the researchers use both absolute and parsimonious fit indices (PGFI, Normed chi-square, PNFI, and χ2 difference test).

To create the TRA model, constraints set the γ13 path to zero. Fitting this model to the overall covariance matrix results in the

Fig. 3. Revised structural model for the TPB (Model 2).

following goodness-of-fit indices: χ2 (39)=106.97, RMSEA=0.1, GFI=0.91, and AGFI=0.86. Fig. 4 shows the results for the TRA.

Table 3 shows the completely standardized (CS) structural coefficients and significance levels for both models. In both, attitude and subjective norm are significant predictors of behavioral intention. However, the structural path from PBC to intention is not significant (even though the fit of the model still meets traditional thresholds) when the fuller model includes PBC.

Table 4 shows a comparison of goodness-of-fit indices between the TPB and the TRA. In general, the goodness-of-fit indices for the TRA model show a slightly better fit than those of the TPB model. No statistically significant difference between the fit of the two models, as evidenced by the χ2 difference test, Δχ2 (1)=0.06 pN0.05, exists. Thus, the TRA model should be retained as this model explains the same amount of variance in intention with fewer parameters to be estimated.

6. Discussion and conclusions

Previous studies (Madden et al., 1992; Taylor and Todd, 1995a; Chang, 1998) find that the TPB, which incorporates the PBC construct, provides a significant better fit than that of the TRA which does not incorporate the same construct. In this study, even though both models fit the covariance matrix relatively well, results from a χ2

difference test reveal no significant difference between the models. The results show the validity of the TRA over the TPB as applied to the domain of adoption of e-commerce among managers/owners of SMEs in Chile. Attitude and subjective norm are significant predictors of

Table 3 Structural coefficients and significance levels for the fitted models.

Relationship TRA coefficients TPB coefficients

A→Intention 0.20⁎⁎ 0.21⁎⁎ SN→Intention 0.73⁎⁎ 0.71⁎⁎ PBC→Intention 0.01

⁎⁎=pb0.01 (1.96).

Table 4 Fit indices for TPB and TRA models.

Model Χ2 RMSEA GFI AGFI PGFI χ2/df PNFI

TPB 106.91 0.094 0.91 0.85 0.53 2.81 0.68 TRA 106.97 0.092 0.91 0.86 0.54 2.74 0.69

297E.E. Grandón et al. / Journal of Business Research 64 (2011) 292–298

intention, whereas PBC is not. Therefore, the model that is more parsimonious (in this case, the TRA) should be retained. In fact, both models explain the same amount of variance on intention (81%). Thus, US researchers need to think carefully about the addition of constructs that may not necessarily reflect the international business world.

These results contradict previous findings regarding the compar- ison between TPB and TRA models (Taylor and Todd, 1995a; Chang, 1998), who found support for the TPB. In making model comparisons, differences in the contexts between the ways researchers conduct studies are important considerations (Taylor and Todd, 1995a). Subjects in the Chang (1998) and Taylor and Todd's (1995a) studies are college students who may have different perceptions about the adoption of IT than managers/owners of SMEs. Perceptions of e- commerce may operate differently in school settings than in workplace settings and the predicted target behavior here is different from the other studies.

The present findings also do not show a positive linear relationship between PBC and the intention to adopt e-commerce. This finding, however, is not too surprising, given the mixed results from previous studies regarding the relationship between PBC and intention. Although Chang (1998) finds that PBC is the strongest predictor of intention, Venkatesh et al. (2003) finds that PBC is a significant predictor of intention only in some of the relationships, and Riemenschneider et al. (2003) report that PBC is not a significant predictor of intention in any of their hypothesized models. The mixed findings could be due to the way the PBC is defined, which may cause inconclusive results in the above mentioned studies, including this one. Armitage and Conner (2001) and Terry and O'Learly (1995) have brought up this criticism. As defined, perceived behavioral control represents perceptions of ease or difficulty of performing the behavior in question (Ajzen, 1991). These perceptions relate to opportunities and available resources to perform the behavior in question.

Another possible explanation for the non-significant relationship between PBC and the intention to adopt e-commerce among managers/owners of SMEs in Chile has roots in cultural differences among the countries in which previous studies have taken place. Nasco et al. (2008) specifically bring up Hofstede in their discussion which provides relevance to discuss the issue again here. The uncertainty avoidance dimension of Hofstede (1997) may be particularly important in the business context represented in this study. Americans exhibit a low uncertainty avoidance index, which means that they are more prone to risk taking and to assume changes. Chileans, on the other hand, exhibit a culture that is less prone to risk taking and may avoid changes. The incorporation of e-commerce brings structural changes and redesign of organizations; determining the required resources (financial, technological, human, etc.) associ- ated with the implementation of e-commerce may be a difficult task. While American managers of SMEs may be comfortable having incomplete information to make the decision to adopt e-commerce, Chilean managers may feel overwhelmed with the uncertainty of not knowing the necessary resources to implement e-commerce, and therefore, may not be able to determine the importance of having these resources available. This fact may explain why PBC was not found to be a significant predictor of the intention to adopt e- commerce among Chilean managers of SMEs in structural tests presented here and in hierarchical results presented in Nasco et al. (2008) and, to a certain extent, why the TRA model is more appropriate than the TPB when explaining e-commerce adoption in this same research context.

6.1. Contributions to academic research and suggestions for future research

By using a sound theory such as the TPB, and by default the TRA, this study intends to fill in the research gap concerning the lack of strong theoretical foundations in IT research involving SMEs in other

parts of the world. One of the major contributions of this research is that the study is the first research that has rigorously applied statistically sophisticated techniques to test the TPB and its variations to predict e-commerce adoption among managers/owners of SMEs in Chile. Results show that, using the TPB, the intention to predict the adoption of e-commerce among managers/owners of SMEs in Chile is approximately 80%. Similarly, the TRA is as good as the TPB and could be even preferred, since the TRA explains the same amount of variance on behavioral intention with less estimated parameters (and consequently, fewer questions to ask managers regarding adoption). Thus, the present results illustrate that, regarding the adoption of e- commerce by managers/owners of SMEs, the TRA could be an excellent and parsimonious model to explain behavioral intention. Researchers may want to replicate this study in similar settings, for example, in other developing countries in South America that share some cultural characteristics with the Chilean one, to corroborate these results. Researchers may also consider testing additional theories to explain e-commerce adoption, (e.g., TAM) or add theories together (e.g., TAM+TRA) to confirm which combination of theories explain better the adoption of e-commerce among manages/owners of SMEs. Finally, other studies should investigate whether the antecedents to adoption intentions change over time.

Future research regarding measurement issues could focus on determining the factor structure of the attitude component of the TPB and TRA. Results from reliability and validity analyses of the attitude construct of this study suggest that two independent constructs could better capture the structure of attitude. For example, as the two items dropped from the attitude construct were negatively-worded, future research could consider positive and negative attitude toward the adoption of the behavior in question. Future research may examine the direct effect of positive and negative attitude on the intention to adopt IT, particularly e-commerce.

6.2. Post hoc modifications

Post hoc analyses on the results of the structural component of the competing models proposed in this study create future research questions. Byrne (1998) emphasizes that model respecification and re-estimation are important for exploratory analysis rather than as a confirmatory one. Modification indices (MI) may be important for determining potential model respecification. Several possibilities exist. For example, a follow-up study involving crossover effects (Taylor and Todd, 1995b) among normative and attitudinal compo- nents of the TRA intention model may be important. Similar to previous work in social psychology and consumer research that allows for crossover effects in the TRA (e.g., Liska, 1984; Oliver and Bearden, 1985), Taylor and Todd (1995b) report a study to predict consumer adoption intention using the TPB. They find that allowing for crossover effects in the TPB results in improvements: crossover effects from normative beliefs to attitude are significant and improve model fit.

In the present study, an examination of modifications for the TPB reveal a high value of 21.48, suggesting a link between one indicator of subjective norm (SN2) to the attitude construct, which clearly represents a crossover effect. Therefore, future research could focus on this type of crossover effect to develop an even stronger theory to predict behavioral intention in other areas of knowledge. Similarly, an examination of modification indices for the TRA model reveals the same pattern of crossover effects, thus, including a link between some of the indicators of subjective norm to attitude may be important for consistency with previous research. For instance, based on previous studies (e.g., Vallerand et al., 1992), Chang (1998) gave support to justify a causal link path linking subjective norm to attitude in the TPB. Consequently, future research could focus on determining the impact of this type of crossover effects on the prediction of intention to adopt IT in a variety of domains. In sum, researchers should continue to take

298 E.E. Grandón et al. / Journal of Business Research 64 (2011) 292–298

established Westernized organizational behavior and managerial theories and examine their applicability to Latin American businesses. Additional goals are to take findings established here (i.e., the importance of the subjective norm construct) and examine whether organizational communications based on the model result in higher levels of e-commerce adoption. For example, the government could use local businesses that have adopted e-commerce as an example to businesses that have not yet adopted, focusing on the normative behavior of other business owners. Finally, studying the other organizational facilitators and inhibitors of e-commerce adoption in Latin American contexts will help the economy of the region continue to grow.

References

Ajzen I. The theory of planned behavior. Organizational Behavior and Human Decision Processes 1991;50:79-211.

Ajzen I, Fishbein M. Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall; 1980.

Armitage CJ, Conner R. Efficacy of the theory of planned behaviour: a meta-analytic review. British Journal of Social Psychology 2001;40(4):471–99.

Bentler PM. On the fit of models to covariances and methodology to the bulletin. Psychological Bulletin 1992;112:400–4.

Bollen KA. Structural equations with latent variables. NY: Academic Press; 1998. Byrne BM. Structural equation modeling with LISREL, PRELIS, and SIMPLIS: basic concepts,

applications, and programming. NJ: Lawrence Erlbaum Associates Publishers; 1998. Chang MK. Predicting unethical behavior: a comparison of the theory of reasoned action

and the theory of planned behavior. Journal of Business Ethics 1998;17:1825–34. CORFO PYME: Un desafío a la modernización productiva. Corporación de fomento a la

producción; 1994. Davis FD. Perceived usefulness, perceived ease of use, and user acceptance of

information technology. MIS Quarterly 1989:319–40. Fishbein M, Ajzen I. Belief, attitude, intention, and behavior: an introduction to theory

and research. Reading, MA: Addison-Wesley; 1975. Fu JR, Farn CK, Chao WP. Acceptance of electronic tax filing: a study of taxpayer

intentions. Information & Management 2006;43:109–26. Harrison DA, Mykytyn PP, Riemenschneider CK. Executive decisions about adoption of

information technology in small businesses: theory and empirical test. Information Systems Research 1997;8(2):171–95.

Hofstede G. Cultures and organizations. Software of the mine. Intercultural cooperation and its importance for survival. New Jersey: McGraw-Hill; 1997.

Khalifa M, Shen KN. Explaining the adoption of transactional B2C mobile commerce. Journal of Enterprise Information Management 2008;21(2):110–24.

Laudon KC, Laudon JP. Management information systems. New Jersey: Prentice Hall; 2004.

Lee SM, Kim I, Rhee S, Trimi S. The role of exogenous factors in technology acceptance: the case of object-oriented methodology. Information & Management 2006;43:469–80.

Liska A. A critical examination of the causal structure of the Fishbein/Ajzen attitude behavior model. Social Psychology Quarterly 1984;47(1):61–74.

Loehlin JC. Latent variable models. An introduction to factor, path, and structural analysis. NY: Lawrence Erlbaum Associates; 1998.

Madden TJ, Ellen PS, Ajzen I. A comparison of the theory of planned behavior and the theory of reasoned action. Personality and Social Psychology Bulletin 1992;18(1):3–9.

Mykytyn PP, Harrison DA. The application of the theory of reasoned action to senior management and strategic information systems. Information Resources Manage- ment Journal 1993;6(2):15–26.

Nasco SA, Grandón EE, Mykytyn PP. Predicting electronic commerce adoption in Chilean SMEs. Journal of Business Research 2008;61(6):697–705.

Nor K, Pearson JM. An exploratory study into the adoption of Internet banking in a developing country: Malaysia. Journal of Internet Commerce 2008;7(1):29–73.

Oliver RL, Bearden WO. Crossover effects in the theory of reason action. Journal of Consumer Research 1985;12:324–40.

Orbell S, Hodgkins S, Sheeran P. Implementation intentions and the theory of planned behavior. Personality and Social Psychology Bulletin 1997;23(9):945–54.

Pavlou PA, Fygenson M. Understanding and predicting electronic commerce adoption: an extension of the theory of planned behavior. MIS Quarterly 2006;30(1):115–43.

Pee LG, Woon IM, Kankanhalli A. Explaining non-work-related computing in the workplace: a comparison of alternative models. Information & Management 2008;45:120–30.

Randall DM. Why students take elective business ethics courses: applying the theory of planned behavior. Journal of Business Ethics 1994;13:369–78.

Riemenschneider CK, McKinney VR. Assessing belief differences in small business adopters and non-adopters of web-based e-commerce. Journal of Computer Information Systems 2001–2002:101–7.

Riemenschneider CK, Harrison DA, Mykytyn PP. Understanding IT adoption decisions in small business: integrating current theories. Information and Management 2003;40:269–85.

Sheppard BH, Hartwick J, Warshaw PR. The theory of reasoned action: a meta-analysis of past research with recommendations for modifications and future research. Journal of Consumer Research 1988;15:325–43.

Taylor S, Todd P. Understanding information technology usage: a test of competing models. Information Systems Research 1995a;6(2):144–77.

Taylor S, Todd P. Decomposition and crossover effects in the theory of planned behavior: a study of consumer adoption intentions. International Journal of Research in Marketing 1995b;12:137–55.

Terry DJ, O'Learly JE. The theory of planned behaviour: the effects of perceived behavioural control and self-efficacy. British Journal of Social Psychology 1995;34:199–220.

Vallerand RJ, Deshaces P, Cuerrier J, Pelletier LG, Mongeau C. Ajzen and Fishbein's theory of reasoned action as applied to moral behavior: a confirmatory factor analysis. Journal of Personality and Social Psychology 1992;62:98-109.

Venkatesh V, Morris V, Davis G, Davis F. User acceptance of information technology: toward a unified view. MIS Quarterly 2003;27(3):425–78.