INTRODUCTION
Structural Equation Modelling (SEM) is a comprehensive statistical method used in testing hypotheses about causal relationships among observed and unobserved (latent) variables and has proved useful in solving the problems in formulating theoretical constructions (Reisinger,1999). Its function have found to be better than other multivariate statistics techniques which including multiple regression, path analysis and factor analysis. Other statistics techniques could not take into consideration due to the interaction effects among depend and independent variables. Therefore, a method that can examine a series of dependence relationships simultaneously helps to address complicated managerial and behavioural issues. SEM also can expand the explanatory ability and statistical efficiency for model testing with a single comprehensive method (Pang, 1996).
Steenkamp and Baumgartner (2000) reflect on the role SEM in marketing modelling and managerial decision making. They discuss some benefits of it. They said that although SEM has potential for decision support modelling, it is probably most useful for theory testing, which is a key phase in developing marketing models [For SEM and LISREL see; Byrne (1998), Cheng (2001), Cudeck et al. (200), Hayduk (1987), Joreskog and Sorbom (2001)].
Applied to data on attitudes, perceptions, stated behavioural intentions, and actual behaviour, SEM can be used to specify and test alternative causal hypotheses. It has been found that, as might be expected, causality is often mutual. The assumption that behaviour is influenced by attitudes, perceptions, and behavioural intentions without feedbacks does not hold up when tested using SEM. These results challenge the assumption, held by some, that stated preference choices can be directly scaled into revealed-preference choice models. It was used path analysis to demonstrate empirical evidence that the causal link from choice behaviour to attitudes is stronger than the link from attitudes to choice behaviour. Subsequent studies using different forms of simultaneous equation modelling showed consistently that attitudes, especially perceptions, are conditioned by choices, while at the same time, attitudes affect choices (Golob, 2001b) . Gärling et al. (2001) explores decision making involving driving choices by using a SEM with latent variables to test links among attitude towards driving, frequency of choice of driving, and revealed presence of a certain type of decision process known as script-based. Golob (2001a) tested a series of joint models of attitude and behaviour to explain how both mode choice and attitudes. Applying Weighted Least Squares (WLS) estimation to a data set from San Diego California, the author demonstrates that choices appear to influence some opinions and perceptions, but other opinions and perceptions are independent of behaviour and dependent only upon exogenous personal and household variables. None of the models tested found any significant effects of attitudes on choice.
Most papers written have focused on the variables explaining the attraction of shopping center choice [ For exammple;Suarez et al.( 2004), Degeratu et al. (2000),Severin et al. (2001) ]. They have always used logit models and random effect model. Degeratu et al. (2000) focus specifically on assessing whether brand names and price have impact on choices online and traditional supermarkets. Severin et al. (2000) investigated use of relatively recent developments in random utility theory to assess the stability over time and space of the preferences underlying retail-shopping choice. They found that good quality, wide selection, good service, nice atmosphere and convenient location were significant choice of retail shopping center model. They noted that high and low prices and latest fashion were not consistently significant in the separate years models. They also showed that convenient location had the largest impact on the shopping center choices.
METHOD
The study has been designed to research factors which consumers consider while choosing shopping centers and to develop a suggestion model for shopping center choice. Beside demographic questions, effective factors determining peoples’ shopping center choice were asked and for 17 items, answers were taken with composed of five Likert-scale (5=very important and very unimportant ). These items are given in Table 1.
Reliability coefficient of questionnaire was calculated as Cronbach =0.79. When the items reduced alpha value were deleted reliability coefficient increased to 0.81. In this study, latent structure is composed of choosing shopping center (E) and explanatory structures are composed of features of materials sold (A), Attitude and behaviour of staff (B), Geographic location of shopping center (C), Easement of Price (D), Regularity at the shopping center (F). The structure, composed of relationship of assumed 5 independent latent variables (A,B,C,D and F) to one dependent latent variable (E) constitute the model to be tested. Hypothesis developed to test the relationship among the latent constructs are given below:
H1 ; There is a significant relationship between choosing shopping center and features of materials sold at the center.
H2 ; There is a significant relationship between choosing shopping center and attitude and behaviour of staff,
H3 ; There is a significant relationship between choosing shopping center and geographic location of the center.
H4 ; There is a significant relationship between choosing shopping center and easement of Price.
H5 ; There is a significant relationship between choosing shopping center and regularity at the shopping center.
FINDINGS AND DISCUSSION
In this study, four models related to latent variables assumed affected to a choose shopping center have been tested by using LISREL computer program with SEM. At first, Model M1, in which all independent variables took place, has been analysed. Analysis results are given in Table 1. When the Table 1 analysis results for M1 are investigated, it is seen that A,B,D and F latent variables are not significant, goodness of fit index are close to acceptable limits and explanatory ability is 52%. Path diagram for M1 is given in Figure 1. Finally, when the M2 results, found by subtracting B, D an F from model, are observed it is seen that A and C parameter estimates are significant and fitness criteria are in the acceptable limits. R2 values of analysed models are calculated as 0.52 and 0.77 respectively. When the best proper model, M2, is observed, 77 % of the dependent latent variable that is choosing shopping center is explained with A and C independent latent constructs. H2, H4 and H5 assumptions for M2 have not been approved. Path diagram for M2 is given in Figure 2, parameter estimates of the model and t values are given in Table 2. Parameter estimates for A-E and C-E relationships in Table 2 and Fig.2 are 0.67 and 0.50, respectively. These coefficients are positive and statistically significant. Analysed models results show that closeness to the address, discount card application, market image and easement of access to the shopping center, respectively form the priority in preference of consumer choice of shopping center. Besides, advertisement of the shopping center and neighbor advice take important role in choice. Findings revealed that behaviour of sales staff at the shopping center and discount cards increase preferability; on the other hand, easement of access and closeness to their addresses take priority in choosing shopping center relative to easement of price.
CONCLUSION
In conclusion, dependent latent variable that is choice of shopping center can be explained with a rate of 77% through independent latent variables i.e. features of materials sold and geographic location of the shopping center. The meaning of unexplained part with 23% is that the consumers choose shopping center considering other factors which are not taken into account in this study. The M2, found the best model in the study, is a suggestion model which depends upon a few amount of data set. It is possible to reach models having high rates by increasing data amount with alternative models.
Table 1- Items related to shopping center choice (Model 1 )
Estimation of parameter t-value
A- Features of materials sold (A)0.341.42
Brand name of materials sold (A1)0.31**3.99
Quality of materials sold (A2)0.46***7.80
Low Prices (A3)0.071.29
Wide selection (A4)0.27**5.09
B- Attitude and behaviour of staff (B)0.010.03
Behaviour of sales staff (B1)0.51***11.3
Geniality of staff (B2)0.53***11.6
C-Geographic location ( C)0.41**3.08
Closeness to the address (C1)0.67***8.64
Easement of access (C2)0.70***9.45
D- Easement of Price (D)0.26*2.06
Payment shape (D1)0.35**5.83
Promotion on selling (D2)0.58**7.30
Discount card (D3)0.69***9.22
F-Regularity (F)-0.06-0.3
Well-organized (F1)0.42***8.45
Moving at the shopping center
without difficulty (F2)0.43***8.36
E-Choosing shopping center (E )
Neighbor advice (E1)0.60
Advertisement (E2)0.59**6.42
Image (E3)0.61**6.55
*p0.05, **p0.01, ***p0.001
Fig.1. Path diagram for M1 Model
Goodness of Fit:NFI: 0.85, NNFI: 0.87, CFI: 0.90, GFI: 0.91, AGFI: 0.87, 2 /df= 2.21
Table 2- Items related to shopping center choice (Model 2)
Estimation of parameter t-value
A- Features of materials sold (A)0.67**4.76
Brand name of materials sold (A1)0.61**5.83
Quality of materials sold (A2)0.25**4.34
Wide selection (A4)0.17*2.79
C-Geographic location ( C)0.50**4.11
Closeness to the address (C1)0.66***7.81
Easement of access (C2)0.71***8.59
E-Choosing shopping center (E)
Neighbor advice (E1)0.53
Advertisement (E2)0.62**6.17
Image (E3)0.64**6.29
*p0.05, **p0.01, ***p0.001
Fig.2. Path diagram for M2 Model
Goodness of Fit:NFI: 0.80, NNFI: 0.79, CFI: 0.87, GFI: 0.94, AGFI: 0.88, 2 /df= 3.42
REFERENCES
Byrne, B.M. (1998). Structural Equation Modeling with LISREL, PRELIS, and SIMPLIS: Basıc concepts, Applications, and Programming, New Jersey : Lawrence Erbaum Associates Publisher.
Cheng, E.W.L. (2001). SEM Being more Effective than Multiple Regression in Parsimonious Model Testing for Management Devolopment Research. Journal of Management Development. 20 (7) , 650-667.
Cudeck, R., Toit, D.S. & Sörbom, D. (2000). Structural Equation Modeling: Present and Future, Scientific Software International Inc.
Degeratu, A.M. , Rangaswamy, A. & Wu, J. (2000). Consumer choice behavior in the online and traditional supermarkets: The effects of brand name, price and other search attributes. Intern. J. of Research in Marketing, 17, 55-78.
Hayduk L.A. (1987). Structural Equation Modeling with LISREL Essential and Advances. The John Hopkins University Press.
Gärling, T., S. Fujii & Boe, O. (2001). Empirical tests of a model of determinants of script-based driving choice. Transportation Research F. 4, 89-102
Golob, T.F., 2001(a). Joint models of attitudes and behavior in evaluation of the San Diego I-15 Congestion Pricing Project. Transportation Research A 35, 495-514.
Golob T.F.,2001(b). Structural Equation Modeling for Travel Behavior Research. UCI-ITS-WP-AS-01-2, Center for Activity Systems Analysis Institute of Transportation Studies University of California, Irvine; Irvine, CA 92697-3600, U.S.A., November 2001, Institute of Transportation Studies University of California.
Joreskog, K. & Sorbom, D. (2001). LISREL 8: User’s Reference Guide, Scientific Software International Inc.
Pang, N.S.K. (1996). School Values and Teachers’ Feelings: a LISREL model. Journal of Educational Administration. 34 (2), 64-83.
Reisinger, Y. & Turner, L. (1999). Structural Equation Modeling with LISREL: Aplication in Tourism., Tourism Management. .20, 71-88.
Suárez, A. , Rodríguez del Bosque, I. , Rodríguez-Poo, J.M. & Moral, I., (2004). Accounting for heterogeneity in shopping centre choice models. Intern. J. of Research in Marketing, 11, 119-129.
Severin V., Louviere, J.J & Finn, A.(2001). The stability of retail shopping choices over time and across countries. Journal of Retailing, 77, 185-202.
Steenkamp, B.E.M. & Baumgartner, H. (2000). On the use of structural equation models for marketing modeling. Intern. J. of Research in Marketing, 17, 195-202.