THE DYNAMICS OF EXPORTING SMEs' INFORMATION BEHAVIOUR Liane J.A. Voerman, Michel Wedel and Peter S. Zwart

Abstract As the environment of European companies changes rapidly, more and more SMEs feel the need to go international. One of the basic questions is how to succeed in an international setting. In this paper, we will present an integral model of export performance, based on an extensive review of almost thirty export performance studies. One of the crucial elements pertaining to export success appears to be the export information behaviour of the exporting SME. In this paper, export information behaviour is measured as the gathering of formal and informal information, at home as well as abroad, and is linked with firm size, national culture, the environment and export performance. In the one-year static model, the information behaviour depends strongly on the firm size and on the country-of-origin (measured as a dummy variable and as national culture through Hofstede's dimensions). Besides, the amount of formal and informal information acquired has a strong link with the international market. Overall, the more turbulent the environment (a combination of the labour market, the supply market, the sales market, and the capital market), the more information is collected. Concerning the link between information behaviour and export performance, even when we control for covariates (firm size, country-of-origin, and the environment) the amount of information gathered positively influences the export performance, with foreign export market research information being most valuable. Next, tentative longitudinal Lisrel models are estimated on two-year-data. The results demonstrate mixed results. Some of the results found in the static models can be found in the longitudinal models, such as the strong impact of changes in the environment and culture on the amount of information collected. Yet, most of the variance in the 1995 export performance measure is accounted for by the same measure in the preceding year. Therefore, these models need further exploration and only tentative conclusions can be drawn. Introduction The environment of European companies changes rapidly, forcing them to broaden their business perspective and to lose the myopic focus on domestic business. The European Union and the democratisation of Eastern Europe are unmistakable opportunities for enterprising companies. Starting negotiations with ten Eastern European countries and with Cyprus, the EU is heading for an internal market of nearly 500 million consumers, reaching from the Atlantic Ocean to the Russian Federation and from the North Pole to the Black Sea. This has several important implications for the companies operating in these countries. Not only do these firms have increased opportunities to market their products and services abroad; their competitors in domestic and foreign markets do as well. Consequently, the competition on their domestic market intensifies, giving many companies a reason to start internationalising. Other important changes include issues such as the shortening product life cycle of most products and services, the globalisation of consumers and co-operation, and the forthcoming implementation of the European Monetary Union. All these factors lead to a more turbulent environment and to increased competition, placing more stringent demands on the product and the companies. These trends will be a very trying challenge for the SMEs: for them to survive a reaction might be necessary. The question is how do SMEs need to react in order to perform successfully on the internationalisation path.

Our paper will address this issue by presenting an integral export success model, derived from an extensive literature review, which deals with export performance studies that appeared between 1988 and 1998. One of the crucial factors determining export performance will turn out to be the export strategy, which includes variables that are related to the formal and informal information behaviour of SMEs. Not isolated in itself, the export strategy of SMEs is partly determined by other internal and external factors, such as the national culture of the SME, the company size, and changes in the firm's environment. Focusing on these parts of the proposed model, we will investigate the static and dynamic relationships between these concepts: national culture, firm size, and environmental changes related to the information behaviour (formal and informal) of firms and their export performance. Towards an Integral Model After a careful review of almost thirty cross-sectional export performance studies, the various variables incorporated by the respective researchers can be grouped according to five conceptual broader constructs: Firm Characteristics, Environment, Export Strategy, Managerial Attitude and Characteristics, and Export Performance (Voerman and Zwart, 1997). By examining the direct and indirect effects found by the studies, an overall picture emerges of a comprehensive model, including the direct and the indirect relationships between the five constructs (figure 1 -- omitted). Some noteworthy remarks emerge from our review. First, there is still no agreement on the measurement of the export performance concept. Although more and more researches use a multiple measure, and the attention for subjective variables is growing, no consensus has been reached yet. In a recent article, Shoham (1998) pleads for a three-dimensional conceptualisation. He states that export performance is a composite outcome of a firm's international sales, which includes three subdimensions: export sales, export profitability, and export growth. Each of the three should include both objective and subjective measures, which are equally important. Second, the complicated nature of the model with all its interrelationships calls for a stronger method of analysis than the widely used regression or discriminant analysis. Third, a call for longitudinal analysis of the data is in its place, as exporting is an ongoing event for most firms. Information Behaviour One of the crucial concepts in the model is the Export Strategy of the SME. This factor contains several variables, among which the use of information and assistance. Quite some literature can be found on the export market information behaviour of SMEs, sometimes distinguishing between formal and informal information (e.g. Johanson and Vahlne 1977, Hart et al. 1994), or, more specifically, between export market research (formal), export market assistance (formal) and export market intelligence (informal) (Souchon and Diamantopoulos 1996). The overall picture that emerges from this literature is that it is beneficial for the exporting SME to collect data and to use it subsequently (Voerman, Wedel, and Zwart 1998a). Next, as figure 1 shows, firm characteristics such as size can affect this information behaviour, as can environmental variables related to the home country and the export market. Unfortunately, only a small amount of literature can be found on the specifics of these variables and information behaviour, although some differences have been found according to firm size (e.g. Hart and Diamantopoulos, 1993), country-of-origin (e.g. Zaheer and Zaheer 1997) and turbulence of the environment (e.g. Wright and Ashill, 1998). Souchon and Diamantopoulos (1996) proposed a conceptual framework of information use in an export setting. In this model they assume, among other relationships, that both organisational factors (such as company size and organisational culture) and the state of the environment (being turbulent or stable) influence the information acquisition (either export market research, assistance or intelligence). This model resembles our proposed integral model and the operationalisation of the model in this paper. Trying to add insights, we investigate

the triangle between the factors mentioned, the information behaviour, and a firm's export performance, both in a static and a longitudinal setting. The empirical part of this paper draws on data gathered through an international mail survey by the INTERSTRATOS group (INTERnationalization of STRATegic orientation of Small- and medium-sized enterprises)1 INTERSTRATOS is a longitudinal research project into the internationalisation of manufacturing SMEs in Europe, covering an annual survey research in seven European countries, i.e. Austria, Belgium, The Netherlands, Switzerland, Norway, Sweden and Finland, and five manufacturing industries. In this research, we define an SME as a private, independent company with less than 500 employees2, resulting in a sample size of 3562 respondents. Survey items of interest are the questions concerning the external (domestic and foreign) information sources respondents consulted, the export sales, export ratio (i.e. export sales as a percentage of total sales), the country-of-origin, the firm size and 12 questions concerning the perception of the respondent with regard to changes in the environment (see appendix 1 (omitted) for a complete overview of all questions). We operationalize firm size both as the number of full time employees and by categorising this number into five categories (0-9 employees, 10-49 employees, 50-99 employees, 100-199 employees, and 200-500 employees). Country-of-origin will be operationalized both as a dummy variable (in the static analyses) and by using two of Hofstede's dimensions, namely Uncertainty Avoidance and Individualism dimensions (Hofstede, 1980). Following the various information types mentioned above, table 1 shows the information sources and the categories they pertain to. The amount of information acquired we operationalized as the number of information sources SMEs consulted, both at home and abroad. Table 1 -- Categorisation of the Domestic and Foreign Information Sources Johanson & Vahlne, Souchon & Hart et al. Diamantopoulos Formal Information

Export Market Research

Export Market Assistance Informal Information Export Market Intelligence

Information sources (at home and abroad) Training institutions Business consultants Credit agencies Chambers of Commerce Research institutions International organisations (e.g. EU) Public promotion fairs Suppliers Customers Export clubs National trade fairs International trade fairs

The 12 questions related to the environment ("How do you see recent changes (last 12 months) from your firm's viewpoint?") were measured on a five-point scale, which (after recoding) ranges from -2 (very negative) to +2 (very positive) with 0 as a neutral point in the middle. The 12 questions pertain to four broader concepts, namely the labour market, the supply market, the sales market and the capital/credit market. Consequently, four new variables were computed, which stand for the total amount of changes, by counting the absolute scores on the three

questions belonging to that specific concept (so, 0 stands for three times a neutral score and, thus, no changes perceived, and 6 stands for three times very negative/positive changes, and, thus, many changes perceived). Regarding the division between informal and formal information, we looked both at the absolute number of sources and the relative emphasis put on either formal or informal sources by computing percentages. All variables are the input for both the static (building upon the 1995-data) and the longitudinal analyses (1994- and 1995-data). Static Analyses The first analyses consider the influence of the firm size (categorised), and national culture on the formal and informal information behaviour, using bivariate analyses of variance3. Next, the influence of the environment on the information behaviour will be analysed using bivariate correlations and multiple regressions. Informal and Formal Information versus Country-of-Origin and Firm Size The country differences are more stringent on the domestic information market, with Finland consulting significantly more formal and informal sources, and Austrian firms using significantly more formal sources than the other countries. Contrary, Austrian firms hardly use any information provider on the foreign market. These findings are reversed for Swedish firms, who use more formal and informal sources than the other SMEs on the foreign market, but lag behind in asking either informal or formal information on the local market. Norwegian SMEs consult significantly fewer sources in all categories. Besides consulting more information sources overall, Finnish firms consider the export market intelligence sources more than average, both at home (55%) and abroad (79%). Contrary, when consulting domestic sources, Austrian firms mainly go to research- and assistance information suppliers. Norwegian firms give more emphasis to formal sources both at home and abroad. Therefore, not only the total number of sources that the respondents consulted within a category differs according to the national culture; the relative importance of each category varies. From a country-of-origin/national culture perspective, dimensions such as uncertainty avoidance and individualism can be related to SMEs' information behaviour. As expected, countries scoring higher on "Uncertainty Avoidance" express a higher need to consult information providers, both at home and abroad. When divided into formal and informal information providers, this higher need only holds on the domestic market. As for the second dimension, more individualistic oriented firms search less (formal or informal) information at home than more collectivistic countries, although they do consult more (formal or informal) information sources abroad. Concerning the relative share of the information types used, the collectivistic nature of firms does not affect the choice on foreign markets, while uncertainty avoiding SMEs turn relatively more often to intelligence sources when searching abroad. Contrary, on the home market more individualistic companies consult relatively more intelligence sources, while more uncertainty avoiding SMEs consult relatively more research and assistance sources. This is consistent with the fact that formal information is known to reduce uncertainty more, leading to a higher reliance on formal information at home, and with the fact that consulting a foreign formal source is a rather large step for an uncertainty avoiding firm, leading to a higher reliance on intelligence sources abroad. Our results also show that the extent of formal information consulting increases significantly with firm size, both on the domestic and on the foreign market. Fewer differences can be found concerning the informal sources. Besides, we see that larger SMEs rely more on formal sources when gathering information on the foreign market. Information Behaviour and the Environment Table 2 presents the results from the analyses with the 12 environmental factors and the information behaviour.

Table 2 -- Regression Coefficients of Environment Variables on Information Sources Independent Labour market

Supply market

Qualified Unskilled Raw Apprentices Dependent Employees Employees Materials

Products Technologies for Resale

Formal info at home Bivariate

-.054**

Multiple

-.052**

-.045**

.067***

.046**

.068***

Informal info at home Bivariate Multiple Formal info abroad Bivariate

.086***

Multiple

.076***

.054**

.076***

-.087*** -.068***

Informal info abroad Bivariate

.076***

.095***

-.136***

Multiple

.056**

.070***

-.101***

-.053**

-.067***

.048**

-.071**

Perc. formal info at home Bivariate Multiple

-.070***

.045*

-.038*

Perc. formal info abroad Bivariate

.054*

.083***

Multiple Independent Sales market Dependent

Local

National

Capital/credit market Financial Advice

Capital Costs

.057**

.059***

-.062**

.092***

.065***

-.051**

International

Finance

Formal info at home Bivariate Multiple

-.070***

Informal info at home Bivariate

.050**

Multiple

-.043**

.053**

Formal info abroad Bivariate

-.079***

Multiple

-.095***

-.040*

.102*** .141***

Informal info abroad Bivariate

-.125***

Multiple

-.137***

Perc. formal info at

-.075***

.088*** .116***

-.061***

home Bivariate

.057**

Multiple

.078***

Perc. formal info abroad Bivariate

.066**

Multiple

.088***

The bivariate entry of the table shows the correlation between the environment variable and the information variable. For instance, the bivariate correlation between the changes on the market for qualified employees and the amount of formal information acquired abroad is .086. This means that the amount of formal information acquired in the export market increases when the changes on the market for qualified employees are considered more positively (remember the coding of the environmental variables, positive scores meant positive perception of changes on the respective market). On the other hand, the bivariate correlation between the changes on the market for raw materials and the amount of formal information abroad is negative (-.087) indicating a more extensive search abroad when the changes on the raw materials market are deemed negative. The multiple entry in the table should be read across the columns. So, in a multiple regression between the amount of formal information acquired abroad as the dependent variable and the various environmental variables as the independents, only positive changes on the markets for qualified employees, negative changes on the market for raw materials, the negative changes on the local sales market, and positive changes on the international sales market induce firms to acquire more information from research and assistance sources on the foreign market. The most striking findings are the relative large impact the changes in the environment have on the information sought abroad, and the large impact of changes on the international sales market. The results show that when the respondent sees the changes on the international market more favourably, he or she starts to collect more informal or formal information abroad. Contrary, if the local market changes favourably respondents do not turn to as many foreign sources as they would when the changes were seen as unfavourable. This, of course, makes common sense. When entrepreneurs see their local market in a pessimistic view, they look around for other possibilities. Accordingly, when the international sales market looks very attractive, they collect more information on this market. Besides, we used the computed overall scores on changes (irrespective of the changes being considered positive or negative) within each concept (labour, supply, sales, and capital market) and analysed whether more changes in the environment would lead to more information acquisition. Table 3 presents the results of these analyses analogous to the previous tables. Again, the amount of changes on the sales market both impacts the amount of information significantly and positively, and impacts the relative importance of formal information at home significantly albeit negatively.

Table 3 -- Regression Coefficients of aggregated Environment variables on Information Sources Independent Dependent

Changes on Changes on Changes on Changes on Labor Market Supply Market Sales Market Capital Market

Formal info at home Bivariate

.117***

.103***

Multiple

.100***

.079***

.114***

.90***

.060***

.066***

Informal info at home Bivariate Multiple

.125*** .120***

Formal info abroad Bivariate

.099***

Multiple

.097***

Informal info abroad Bivariate

.167***

Multiple

.176***

Perc. Formal info at home Bivariate

-.082***

Multiple

-.091***

.060**

Perc. Formal info abroad Bivariate Multiple Information Behaviour and Export Performance Subsequently, the impact of the factors on the relationship between export performance (measured through export sales) and information behaviour was examined through partial correlations. The inclusion of the firm characteristics in the relationship between export market information and performance leads us to two conclusions. First, though the amount of export market information and the export performance are positively correlated in all cases, intervening characteristics such as firm size and national culture influence this association. Ignoring these

covariates might lead to incorrect conclusions in the assessment of the importance of export market information for export performance. On the other hand, the inclusion of the environmental variables has little to no effect on the parameters. Next, not all information sources influence the export sales to the same extent, with export market research information received on the foreign market having the strongest correlation with export sales. The longitudinal model After examining the static relationships between the concepts of interest, a longitudinal approach will be taken, using the 1994 and 1995 INTERSTRATOS data. To get some preliminary insights into these relationships we started with two-year data. As mentioned before, the complicated direct and indirect effects between the various concepts call for a powerful method of analysis. Structural equations modelling (SEM) has several properties making it the appropriate method to tackle a model with the kind of direct and indirect effects as we found (see figure 1 -- omitted). Besides, SEM is the perfect technique for the estimation of models with latent variables, measured by various manifest variables. The Basic Model. As one of the SEM approaches, LISREL gives us the possibilities to estimate the basic model with both information and export performance as pictured in figure 2 (omitted). The manifest and latent variables to be used in this analysis and subsequent analyses are reflected in table 4. Table 4 -- Latent and Manifest Variables in the Study Constructs

Manifest Variables

Firm Size

Number of employees

Culture

Score on Uncertainty Avoidance Score on Individualism

Labour Market

Supply Market

Sales Market

Credit Market

Qualified employees Unskilled workers Apprentices Raw materials Products for resale Technologies Local market National market International market Finance Financial advice Interest and other capital costs

Information Behaviour

Total number of informal information sources Total number of formal information sources

Export Performance

Export sales Export ratio

Underlying hypotheses of the model are that the firm size, the national culture, and the environment all influence the information behaviour. This information behaviour has a positive impact on the export performance within the same year, but there is no lag in the effects of this data. The success of the previous year has a strong linkage with the success in the following year, but will also influence the amount of information providers consulted in the following year. This amount will not be independent of the information behaviour in the previous year, therefore a relationship is assumed (see figure 2). Sample Size One of the disadvantages of using two-year data is the decrease in sample size4. Not all firms used in the static analyses also participated in the 1994 survey, and not all firms filled in all questions on the questionnaire. Especially the variables 'total number of informal information sources', 'total number of formal information sources', and 'export sales' have many missing values. Next, a proper analysis of all effects requires a single respondent to have no missing values on the variables of interest (listwise deletion). Therefore, after imputation of missing values on the base of their values in the other year(s), the total sample size for the longitudinal analysis is 191 SMEs. PRELIS Calculations LISREL assumes the variables to be continuous and multivariate normal. However, in practice researchers often have to do with ordinal variables, measured by a five- or seven-point Likert scale, as are our variables pertaining to the environment. To tackle this problem, the program PRELIS allows you to calculate a correlation matrix with polychoric correlation coefficients (both ordinal variables) and polyserial correlation coefficients (between ordinal and continuous variables). Jöreskog and Sörbom (1988) found in a Monte Carlo study that the polychoric correlation is the best approximation for the real correlation between two ordinal variables. Besides, Quiroga (1992) found that the polychoric correlation proved to be best even in the case of a skewed bivariate distribution. Therefore, with the raw data from the SPSS-file as input PRELIS calculated a correlation matrix, which serves as the subsequent input in the LISREL analyses. Although a polychoric correlation matrix demands you to use the Weighted Least Squares (WLS) method, being a distribution-free method, for the analysis, there are several practical problems with this method, such as the requirement of a large data set. lt has been shown that in cases of a relatively small sample size it could even be better to use the Maximum Likelihood (ML) method than the WLS method (Boomsma, 1998). LISREL Estimations Therefore, after calculating the correct correlation matrix in PRELIS, we used the ML method in LISREL to analyse the relationships. The LISREL output gives you various statistics to determine the overall fit of the proposed model. We will use the chi-square value (χ2)5, the Root Mean Square Error of Approximation (RMSEA)6, the goodness-of-fit statistic (GFI)7 and the adjusted goodness-of-fit statistic (AGFI)8 to determine the fit of the various models. Furthermore, we will deal with the face validity of the parameters and the statistical significance of the parameters. Table 5 presents the results of the various models, the bold figures meaning a good fit according to the specific goodness-of-fit statistic. First, we estimate the basic model (see figure 2), followed by the basic model extended with Culture, Firm Size, Environment, and Changes in Environment, respectively, with these factors loading on Information in 1994 and in 1995. Besides, a model with just Information and Culture (Firm Size, Environment, and Changes in Environment, respectively) will be estimated to examine the effects of the respective factors on the acquisition of information without the influences of export performance incorporated in the model.

Table 5 -- LISREL Models and the Goodness-of-Fit Statistics Model

χ2 (p-value) RMSEA GFI AGFI

Basic

.42

.01

.99

.95

Basic + Culture

.01

.06

.96

.91

Basic - Success + Culture

.00

.14

.97

.85

Basic + Firm Size

.00

.19

.88

.70

Basic - Success + Firm Size

.31

.03

.99

.96

Basic + Environment

.00

.10

.76

.71

Basic - Success + Environment

.00

.10

.79

.73

Basic + Changes in Environment

.18

.03

.94

.90

Basic - Success + Changes in Environment

.06

.04

.95

.91

The main conclusions from the models estimated are listed in this section. First, it turns out that the relationships running from Information to Success are not significant. Most of the variance in the Success measure in 1995 is explained by the Success concept in the previous year. The amount of information in 1994 does affect the information acquired in 1995 positively, as does the export performance in 1994. This can be explained by habits and by the costs of collecting information, making it easier to collect information when your export performance was higher. The fit of this basic model was very good, considering all four statistics in table 5. The Extended Models Including the factors Culture, Firm Size, Environment, or Changes in Environment does not change the parameters significantly. Therefore, for reasons of simplicity, we can look at the basic model without the Success factor and with the respective factor. We find that Culture, measured by 'Uncertainty Avoidance' (positive loading) and by 'Individualism' (negative loading), influences the amount of information positively, confirming the static results. The parameters are all significant, have face validity, and the model has good fit according to the GFI and the AGFI. When the Success concepts are integrated in the model, the RMSEA also falls below the .10 and indicates a good fit. Including Firm Size in the basic model does not improve the model well. The sign of the parameters have face validity: in both years a larger firm collects more information, although part of the variability in the Information concept in 1995 is accounted for by the variability in the Information factor in 1994. Unfortunately, the parameters in this model do have face validity, but are not significant despite the good fit statistics. The models with the four environment concepts both have bad fit statistics. Only the RMSEA has a just acceptable value of .10, but the other three statistics indicate a bad fit between the model and the data. Although we cannot attach to much meaning to the outcomes of these models, we can see some things on the parameters. Only a few of the parameters are significant, indicating a significant and positive impact from positive changes on the supply market on information acquisition, while a positively changed credit market only influences the amount of information acquired in the first year. These findings do partly coincide with the static results, but the importance of the sales market disappears in the longitudinal model. Lastly, the model is estimated with two concepts representing the total

changes in the environment on all four submarkets. These models both have a good fit and indicate the importance of the amount of changes in the environment for the entrepreneur to consider information sources. Overall, we can say that the amount of information does not affect success, but does affect the information requested in the following year. This information is also highly influenced by the success in the previous year, as is the success in the following year. The more individualistic firms are, the less information they require, while more uncertainty avoiding firms require more information. The conclusions on firm size and the four environmental concepts should be considered with care, as he parameters are not significant and the fit is very bade, respectively. Contrary, the results on the changes in the environment are very good and indicate a turbulent environment to induce entrepreneurs to turn to more information providers. Conclusions and Future Research With this study we tried to improve our knowledge on the interface between export performance, information behaviour and firm or environmental characteristics, both in a static and a longitudinal way. The static results show strong influences from firm size, national culture and the environment on the information behaviour of SMEs. This can serve as an input for government as well as information providers in determining their policy. Especially changes on the local and international market induce entrepreneurs to collect more formal information on the foreign market. In the longitudinal analyses the influences form information on export performance were not significant. As we can see in the integral export performance model (fig. 1), information behaviour is just one of the various components of which export strategy consists. A next step will be to include more strategic behavioural variables into the model and see if there is a link between export stragey (including information behaviour) and export success. Another step will be to incorporate all factors used in this study into the basic model simultaneously, to investigate their integrated effects on information and export performance. Lastly, more years could be incorporated in the models. Due to missing cases a large part of the sample was not useful. Better imputation methods or the inclusion of other variabels instead of the variables that caused so many cases to be deleted could be a solution. Unfortunately this is not a very plausable option, since the Interstratos survey does not consist of many other variables that can serve as a proxy for information behaviour or export performance. A last remark concerns our focus on the European market. Most of the literature on export market information and export performance focuses on the US market (see also Diamantopoulos et al., 1993), while this study might contribute to a wider understanding of the subject by looking into the information behaviour of European firms. Compared with their US counterparts, European firms operate in a much smaller national environment and are, therefore, sooner inclined to consider their chances on the other side of the border. References Boomsma, A. (1998). Reader coming with the PhD-course Research Methodology, Groningen. Diamantopoulos, A., B.B. Schlegelmilch, and K.Y.K. Tse (1993). "Understanding the Role of Export Marketing Assistance: Empirical Evidence and Research Needs," European Journal of Marketing 27 (4), 5-18. Johanson, J. and J.-E. Vahlne (1977). "The Internationalization Process of the Firm -- A Model of Knowledge Development and Increasing Foreign Market Commitments," Journal of International Business Studies 8 (1), 2332.

Jöreskog, K.G. and D. Sörbom (1988), PRELIS: A Program for Multivariate Data screening and Data Summarization. A Preprocessor for Lisrel. Mooresville: Scientific Software. Hart, S.J. and A. Diamantopoulos (1993). "Marketing Research Activity and Company Performance: Evidence from Manufacturing Industry," European Journal of Marketing 27 (5), 54-72. Hart, S.J., J.R. Webb and M.V. Jones (1994). "Export Marketing Research and the Effect of Export Experience in Industrial SMEs," International Marketing Review 11 (6), 4-22. Hofstede, G. (1980). Culture's Consequences: International Differences in Work-related Values. Beverly Hills: Sage. Quiroga, A.M. (1992). Studies of the Polychoric Correlation and other Correlation Measures for Ordinal Variables, doctoral thesis, Uppsala: Uppsala University. Sharma, S. (1996). Applied Multivariate Techniques. United States: John Wiley & Sons, Inc. Shoham, A. (1998). "Export Performance: A Conceptualization and Empirical Assessment," Journal of International Marketing 6 (3), 59-81. Souchon, A.L. and A. Diamantopoulos (1996). "A Conceptual Framework of Export Marketing Information Use: Key Issues and Research Propositions," Journal of International Marketing 4 (3), 49-71. Voerman, L. and Zwart, P.S. (1997). The Export Performance of Small- and Medium-sized Enterprises: Towards An Integral Model. Small Business Research Memorandum (02). Groningen: Small Business Center. Voerman, J.A., Wedel, M., and Zwart, P.S. (1998a). The Export Market Information Behaviour of SMEs: The Influence of Firm Characteristics. Paper presented at the 43rd ICSB conference Singapore. Voerman, L., Wedel, M. and Zwart, P.S. (1998b). The Triangle Between Firm Characteristics, Export Market Information, and Export Performance. Small Business Research Memorandum (01). Groningen: Small Business Center. Wright, M. and N. Ashill (1998). "A Contingency Model of Marketing Information," European Journal of Marketing 32 (1/2), 125-144. Zaheer, S. and A. Zaheer (1997). "Country Effects on Information Seeking in Global Electronic Networks," Journal of International Business 28 (1), 77-100. About the Authors Liane J.A. Voerman, Michel Wedel, and Peter S. Zwart Department of Marketing and Marketing Research University of Groningen PO Box 800 9700 AV Groningen The Netherlands

Contact person Liane J.A. Voerman Email: [email protected] 1

The international research group known as INTERSTRATOS is formed by J. Hanns Pichler, Christian Lettmayer, Erwin Fröhlich and Peter Voithofer (Austria), Rik Donckels, Ria Aerts and Jan Degadt (Belgium), Antti Haahti and Petri Ahokangas (Finland), Yvonne Prince, Peter Zwart and Liane J.A. Voerman (the Netherlands), Per-Anders Havnes and Arild Saether (Norway), Håkan Boter and Carin Holmquist (Sweden), Margrit Habersaat and Hans Jobst Pleitner (Switzerland).

2

This is a somewhat broader range than the usual definition in the EU (250 employees).

3

Part of these results can be found in Voerman, Wedel, and Zwart (1998b).

4

Besides, to be able to extend these analyses to five-year data and to be able to compare the models found, only those firms are included that participated during five years. 5

The chi-square value indicates the fit between the model and the data. If it is significant, the model does not fit the data very well. However, this statistic is very sensitive to non-normality and large samples. 6

The Root Mean Square Errors of Approximation is one of the most informative criteria and tests whether the model fits the population covariance matrix, if it were available. Values less than .05 indicate good fit, .08 to .10 indicates mediocre fit and larger than .10 indicate poor fit.

7

The goodness-of-fit statistic compares the model with no model at all. The higher this statistic, the better the fit of your model. Sharma (p. 159, 1996) indicates a cut-off point of .90. 8

Same as the GFI, but adjusted for the degrees of freedom. The higher this statistic the better the fit of your model. Sharma (p. 159, 1996) indicates a cut-off point of .80.