Professional Documents
Culture Documents
An Exploratory Study of Trade Show Formation and Diversity
An Exploratory Study of Trade Show Formation and Diversity
[subscription number 918910197] Publisher Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 3741 Mortimer Street, London W1T 3JH, UK
Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t792303971
Jianan Wu a; Gary L. Lilien b; Aniruddha Dasgupta c a E. J. Ourso College of Business, Louisiana State University, Baton Rouge, LA b Smeal College of Business, The Pennsylvania State University, University Park, PA c Massachusetts Maritime Academy, Buzzards Bay, MA
To cite this Article Wu, Jianan, Lilien, Gary L. and Dasgupta, Aniruddha(2008) 'An Exploratory Study of Trade Show
Formation and Diversity', Journal of Business To Business Marketing, 15: 4, 397 424 To link to this Article: DOI: 10.1080/15470620802325617 URL: http://dx.doi.org/10.1080/15470620802325617
1547-0628 1051-712X WBBM Journal of Business-to-Business Marketing, Marketing Vol. 15, No. 4, October 2008: pp. 140
ABSTRACT. Purpose: To analyze the objectives of trade show participants and to link those objectives to trade show formation and diversity. Methodology/Approach: We conduct an exploratory analysis of trade show formation and diversity and link them to the differences in several seller and buyer interests across industries. We used data collected from a single large trade show to examine the nature of buyer and seller participation goals using multivariate statistical analysis. We then used a data set collected from the entire trade show industry to examine the trade show diversity via an econometric model. Findings: We find that higher selling and buying propensity is linked to more vertical shows whereas higher breadth of product interests is linked
Jianan Wu is associate professor of marketing at the E. J. Ourso College of Business, Louisiana State University, Baton Rouge, LA (E-mail: jiananwu@lsu.edu). Gary L. Lilien is distinguished research professor of management science at Smeal College of Business, The Pennsylvania State University, University Park, PA (E-mail: GLilien@psu.edu). Aniruddha Dasgupta is associate professor of international maritime business at the Massachusetts Maritime Academy, Buzzards Bay, MA (E-mail: adasgupta@maritime.edu). We gratefully acknowledge financial support from the Institute for the Studies of Business Markets (ISBM) of Penn State University. We thank Hans Baumgartner, Kalyan Chatterjee, Srinath Gopalakrishna, and Arvind Rangaswamy for many useful comments on an earlier version of this paper and Exhibit Surveys, Inc. for providing the data used in the empirical study. Address correspondence to: Jianan Wu, E. J. Ourso College of Business, Louisiana State University, Baton Rouge, LA 70803 (E-mail: jiananwu@lsu.edu). Journal of Business-to-Business Marketing, Vol. 15(4) 2008 2008 by The Haworth Press. All rights reserved. doi:10.1080/15470620802325617
397
398
to more horizontal shows. We also find that a highly innovative industry is often linked to more horizontal shows. Originality/Value/Contribution: Although trade shows are critical elements of the communications mix in business-to-business marketing, very little is known about how a trade show is formed and why two different types of shows (i.e., horizontal and vertical) coexist in the industry. This is the first study that investigates these issues. Our work should help guide trade show organizers diagnose and improve the appropriate mix of shows in a given industry and help trade show participants select better shows to attend, depending on their show objectives and selling/buying interests.
Downloaded By: [Romanian Ministry Consortium] At: 19:51 2 March 2010
Trade shows or industrial exhibitions constitute an important element of the communication mix in the business marketplace. In terms of expenditure allocations, they make up the third largest component of the marketing communication mix behind direct marketing and business magazine advertising (Pinar, Rogers, and Baack 2002). In the United States alone, the trade show industry annually hosts over 12,000 shows, involves over two million exhibitors and over 100 million attendees (CEIR 2001), and generates over 50 billion U.S. dollars (Marken 2004). While the industry has seen some consolidation in recent years, growth in interest and participation remains high (Donberg 2006; The Industrial Robot 2006). The widespread prevalence of this form of market/meeting place in the United States seems representative of the world at large. For instance, Hansen (1996) reports that there were 169 international shows held in Germany with 124,000 exhibitors and over 10 million attendees, as well as 220 international shows organized in the UK with 61,000 exhibitors and 4.6 million attendees. In business markets, two types of trade shows are formed based on market coverage: vertical shows and horizontal shows (Gopalakrishna and Lilien 1994). The nature of the show audience in terms of product interest, buying/selling plans, the time horizon of strategic consideration, and so on, can differ greatly by type of show. Typically, a vertical show involves a narrow range of products and attracts visitors specifically interested in those products. In contrast, a horizontal show usually involves a much broader range of products and a more diverse audience. For example, attendees at the Association of Operating Room Nurses show, a vertical show, are almost all operating room nurses, and exhibitors display products
399
that are used almost exclusively in the operating room. A typical horizontal show is the National Design Engineering Show, in which firms demonstrate products ranging from mechanical components, electrical and electronic components, plastics, elastomers, to CAD/CAM systems (Gopalakrishna and Williams 1992). It seems curious that vertical and horizontal shows coexist in many industries and that the proportion of horizontal shows formed varies widely across industries. Table 1 details the number and proportion of vertical shows and horizontal shows for different industries in a sevenyear span (19851991). There appears to be no simple pattern here: why are 319 of the 322 shows in the communications industry vertical, 61 of the 64 shows in the food processing and distribution industry horizontal while the chemical industry splits almost evenly with 20 vertical and 23 horizontal shows? Even the status of existing shows evolve, with some merging and becoming more horizontal and others splitting into more narrow shows. For instance, Conexpo-Con/Agg 96, the largest U.S. trade show in the construction industry, resulted from a merger of Conexpo (owned by the Construction Industry Manufacturers Association), the largest construction show, and Con/Agg (owned by the National Aggregates Association and the National Ready Mixed Concrete Association), one of the premier vertical shows specializing in concrete and aggregates in the United States. The new horizontal show attracted more than 1,250 exhibitors and 100,000 attendees, and covered 1.25 million square feet of exhibiting area, making it bigger than the last Conexpo and Con/Agg combined (Show Daily of Conexpo and Con/Agg, 96, 18). On the other hand, Comdex, the largest (horizontal) computer show in the United States, has seen threats by key exhibitors to split and form more narrowly focused shows (Calton 1994, A1). The consolidation and separation of trade shows is often strategic. For example, in 2005 the merge of Waste Expo and Wastecon continued to make news in the waste industry. Yet the two shows remain separate entities (Johnson 2005). This diversity and structural evolution of show type has seen little or no academic attention to date. Most trade show studies emphasize saleslead generation for exhibitors as the key goal, providing little rationale for the formation of horizontal shows (Gopalakrishna and Williams 1992; Gopalakrishna and Lilien, 1994; Dekimpe et al. 1997). These empirical studies assume that lead generation is a dominant objective for exhibitors and measure the performance of the show by how efficiently trade shows generate such leads. That performance measure suggests that because
TABLE 1. The distribution of trade show types for top 21 industries (19851991)
Industry Number of Vertical Shows 379 319 3 90 80 3 52 49 48 3 20 0 32 0 15 0 6 21 17 0 15 1,152 441 0 0 44 23 33 1 30 10 24 15 0 0 15 0 49 48 47 43 33 33 30 25 24 21 21 17 15 15 1,593 93 4 0 61 7 96 94 80 64 59 78 3 457 322 Number of Horizontal Shows Total Number of Shows Proportion of Horizontal Shows 17% 1% 97% 4% 0% 95% 12% 0% 0% 94% 53% 100% 3% 100% 40% 100% 71% 0% 0% 100% 0% 28%
400
Frequency
50100 shows
Engineering Medical and health care Housing Food processing and distribution Electrical and electronics
1550 shows
Paint Nursing Automotive and trucking Chemical Plastics Radio, TV, and cable Energy Building and construction Restaurants and food service Manufacturing Photographic Welding Packaging Education
Total
Source: Exhibit Surveys Inc. The industry classification used here are in keeping with the Trade Show Bureaus own classification scheme. Industries with fewer than 15 trade shows have been omitted from this table because of the instability of the proportions when the total number of shows in the industry is that low.
401
vertical shows outperform horizontal shows in terms of selling efficiency, there should be little economic rationale for the formation of horizontal shows. Research to date has not studied why the proportion of show types vary across industries, which is the key question we explore here. To seek a coherent answer to this question requires an investigation of the actors and factors involved in the trade show formation process. Our exploratory analysis suggests that trade show participants selling/ buying propensity and breadth of product interests are closely correlated with the type of trade show formed: the higher the selling/buying propensity the more vertical shows we observe whereas the higher the breadth of product interests the more horizontal shows. Our analysis provides a mechanism to identify those product or market areas that appear either underserved or overserved by one or another type of show. We proceed as follows. In the next section we identify some important variables that theory and logic suggest should be linked to the diversity of trade shows across industries. Next we use data collected from a leading trade show research firm to empirically investigate our propositions. We conclude by discussing the theoretical and managerial implications of our results and the potential for future research in this area.
402
propensity (for attendees and exhibitors respectively), we should expect these ratios to vary from industry to industry. And in some industries such as energy that rely on numerous technologies and see broad applications, participants may be interested in a broad range of products while in others (e.g., paint) those interests are likely to be much narrower. Variation in both these sets of participant goals may be expected to affect the type of trade shows that are formed.
Exhibitors may have different selling propensities when exhibiting at a trade show. Kerin and Cron (1987) examined exhibitors performance using factor analysis and found two categories of roles that trade shows play in the exhibition industry: some exhibitors have objectives that are primarily sales oriented, such as generating leads, while others have nonsales-oriented objectives such as gathering competitive intelligence and enhancing company image. Following Kerin and Crons approach, Hansen (1999) proposed a two-dimensional framework on exhibitors trade show performance: an outcome-based dimension that includes the salesrelated activities, and a behavior-based dimension that includes nonsalesrelated activities, such as information-gathering activities, image-building activities, motivation activities, and relationship-building activities. Hansen (1999) also proposed a dual-motive theory in which selling and nonselling functions may exist not only for exhibitors but also for attendees. Using data collected from the seafood exporting industry of Norway, Hansen (1999) empirically demonstrated general support for his dual-motive theory. Consistent with Hansens (1999) findings, attendees have different buying propensities as well. The results from a Trade Show Bureau (1994) survey demonstrates that some attendees have immediate purchase needs while others may use trade shows to gather new product information that may be useful in guiding future purchases. Godar and OConnor (2001) find further support for this perspective, segmenting trade show attendees motives into two groups: short term (e.g., to make a buying decision or to confirm a prior decision) and long term (e.g., to reinforce contacts, develop contacts, and gather information for future purchase decisions). Exhibitors and attendees prioritize the importance of their information needs when attending a show, and should look for the best return on that information-exchange investment. Exhibitors with low selling propensity
403
should seek to spread the word broadly; attendees with low buying propensity should seek the broadest base of information collection possible. Hence, situations with low selling/buying intensity by exhibitors or sellers should favor horizontal shows over vertical shows as efficient information exchange venues. Therefore, we have: Proposition 1. A lower selling or buying propensity for exhibitors or attendees will be positively correlated with the number of horizontal shows.
Downloaded By: [Romanian Ministry Consortium] At: 19:51 2 March 2010
404
Although it seems intuitive that the higher breadth of product interests by either exhibitor or attendee should make horizontal shows more appealing, the size and possible congestion at a large, heavily attended and diverse show can make such a show inefficient. Conexpo and Con/ Agg 96, a horizontal show formed by merging two vertical shows, illustrates this issue. In the merger decision a major issue for the Con/Agg team was their concern that exhibitors of specialized concrete and aggregates equipment might get lost in a large general construction forum. Robert A. Gale, representing the Con/Agg team, said, The decision on product concentration area was a key. That helped us [on the Con/Agg side] make the decision about 1999. We realized product concentration areas were essential . . . not just for concrete and aggregates suppliers but for many other product categories too (Show Daily of Conexpo and Con/ Agg, 96, 18). This inefficiency of horizontal shows can be critical when the disparity of breadth of product interests between exhibitors and attendees is high. To see the rationale, assume there are two products X and Y. Assume that while all attendees are interested in both products (i.e., a high breadth of product interest for attendees), some attendees (denoted by AX) are more interested in X while others (denoted by AY) are more interested in product Y. Assume also that the exhibitors have a narrow breadth of product interest, leading to a high degree of disparity of breadth of product interests between attendees and exhibitors. Suppose also that some exhibitors (denoted by EX) are only interested in exhibiting X while others (denoted by EY) are only interested in exhibiting Y. Other things being equal, it will be more efficient if EX (or EY) can identify AX (or AY) with less effort (e.g., saving valuable time at a trade show). In a horizontal show with both products X and Y, all EX, EY, AX, and AY will attend. In two vertical shows with products X and Y separated, EX and AX will attend one show while EY and AY will attend the other. Exhibitors (EX or EY) at the horizontal show must incur extra effort in screening out their marginal attendees (AY or AX), effort avoided in vertical shows through self-selection. This inefficiency leads to the separation of a horizontal show into one or more vertical shows where the scanning costs for exhibitors and attendees will be smaller. Therefore, we have: Proposition 3. A greater difference in breadth of product interests between exhibitors and attendees will be negatively correlated with the number of horizontal shows.
405
Industry-Specific Correlates
We discussed trade show specific correlates in trade show formation. However, there are industry-specific correlates in trade show formation as well, a critical one being technological innovativeness.
Technological Innovativeness
One of the fundamental industry characteristics is technological innovativeness. The degree of technological innovativeness affects needs and wants of exhibitors and attendees. When the technology lifecycle is short, it is critical for all actors to stay abreast of the latest developments not only within the core industry but also in adjacent industries into which they can expand or from where future competitors may emerge (Zook and Allen 2001). A cross section of industries (see Table 1) will allow us to examine the effect of technological innovativeness. Specifically, for industries with higher levels of technological innovativeness (which we refer to as technological turbulence), more frequent and broader environmental scanning is necessary to stay abreast of current and future product developments. Therefore, we have: Proposition 4. Higher levels of industry technological innovativeness will be positively correlated with the number of horizontal shows. These ideas and our propositions are summarized graphically in Figure 1. (We discuss the possible effect of other industry-specific correlates further on in the discussion section.)
406
FIGURE 1. A conceptual framework showing the drivers of trade show formation and diversity.
Technological Innovativeness
Low
Technological Innovativeness
High
High
High
Note: This figure suggests that two sets of parties, exhibitors and attendees, can vary in their trade show participation objectives along two sets of objective dimensions (i.e., breadth of product interest and buying and selling intensity) when influencing the formation of the type of trade show. The impact of that variation depends on industry structure (i.e., technology innovativeness).
framework. This analysis, while preliminary in nature, allows us to see if the conceptual constructs on selling/buying propensity and breadth of product interests are empirically supported and may also be used to support further theory development. We next test our Propositions on the drivers of trade show diversity derived from our conceptual framework.
Technological Innovativeness
High
407
(B) Attendees (Sample size = 281, Mean = 5.79) 16.0% 14.0% 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 0.0% Breadth of Product Interest 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Breadth of Product Interest
(A) shows the distribution of number of product types exhibited by exhibitors within the 16 product types and (B) shows the number of product types of interest to attendees, indicating significant cross-product interests for both exhibitors and attendees at this show, but with the breadth of product interests for exhibitors smaller than those for attendees.
TABLE 2. Mean expert evaluations of various characteristics of both the trade shows and the associated industries noted in Table 1
Technological Innovativeness (Industry) (TECH) 6.58 6.32 5.00 5.08 2.66 3.11 5.58 2.26 2.37 3.26 3.71 3.82 3.53 2.61 2.53 2.45 3.79 4.26 2.32 3.03 2.76 3.67 1.33 4.05 1.51 4.29 1.09 3.45 1.25 6.00 5.00 2.00 5.00 5.00 2.50 3.00 6.00 3.00 2.00 3.50 3.00 6.00 3.00 3.00 2.00 3.00 6.00 5.00 5.00 6.00 6.00 6.00 3.00 4.00 4.50 4.00 5.00 3.00 5.00 3.00 4.00 4.00 4.00 3.00 5.00 2.00 5.00 5.00 6.00 4.50 4.00 3.00 2.00 6.00 2.00 2.50 4.50 2.00 2.00 2.00 4.00 3.50 4.50 2.00 5.00 5.00 4.50 4.00 2.50 3.00 4.50 4.00 Buying Interest (Attendee) (ATBUY) Selling Interest (Exhibitor) (ATSELL) Breadth of Product Interest (Attendee) (ATPROD) Breadth of Product Interest (Exhibitor) (EXPROD) 4.00 3.00 5.50 2.00 3.50 4.50 2.00 3.00 4.00 4.50 4.00 4.50 4.00 5.50 3.00 5.00 4.00 2.50 3.00 4.00 2.00 3.69 1.07
408
Industry
Computers and computer applications Communications Engineering Medical and health care Housing Food processing and distribution Electrical and electronics Paint Nursing Automotive and trucking Chemical Plastics Radio, TV, and Cable Energy Building construction Restaurants and food service Manufacturing Photographic Welding Packaging Education
Mean SD
Note: Column 1 (TECH) is the mean from 17 expert evaluations while the other columns represent the consensus judgment after two rounds from our two trade show research experts.
409
We conducted an exploratory factor analysis for the exhibitors objectives and report the results in Table 3. Our results are largely consistent with previous studies (e.g., Kerin and Cron 1987). The factor analysis reveals a two-factor structure of the exhibitors objectives: selling factor and nonselling factor. Objectives such as Number of Leads, Quality of Leads, Introducing New Products, and Selling at the Show are more heavily loaded on the selling factor while objectives such as Improving Company Awareness/Images, Enhancing Corporate Morale, Identifying New Prospects, Market Testing New Products, and Gathering Competitive Information are more heavily loaded on the nonselling factor. The other three objectives are more evenly loaded on both factors. The twofactor model explains 52% of the total variance. In the attendee survey, attendees were asked which of the following reasons explained why they attended. See new products/developments Fact finding for future purchases Make a purchase Attend seminars/association meetings
410
Note: The Eigen values of the factor analysis are 4.68, 1.61, 0.97, 0.84, 0.73, 0.65, 0.58, 0.45, 0.44, 0.37, 0.35, 0.29. The minimum Eigen value criterion suggests a two-factor pattern.
To see specific company(s)/product(s) To find products to represent Solve a problem Network with peers Obtain technical or product specifications
As these are binary (Yes-No) data, a standard approach is inapplicable. An appropriate approach is to assume that there is a latent continuous variable xi* and a threshold constant t (e.g., Bock and Lieberman 1970; Christofersson 1975; Muthen 1978) for each observed index xi , in which xi = 1 if and only xi* > t . Then, the latent factor analysis model can be written as x* = l F + e , which has the same mathematical form as the usual common factor analysis model, and e is assumed to have a multivariate normal distribution with mean zero. Because the latent variable x* is unobservable and the observed variable x is dichotomous, the estimated correlation matrix will be the tetrachoric correlation matrix. There are several estimators developed for this model. Bock and Lieberman (1970) used an MLE estimator and Christofersson (1975) and Muthen (1978) developed a GLS estimator. We use Christoferssons approach via PRELIS of LISREL and report the tetrachoric correlation matrix in Table 4a.
411
a. The tetrachoric correlation matrix of attendees objectives See new products/developments 1.00 Fact finding for future purchases 0.19 1.00 Make a purchase 0.11 0.12 1.00 Attend seminars/association 0.16 0.19 0.22 1.00 meetings To see specific companies(s)/ 0.09 0.10 0.09 0.18 1.00 product(s) To find products to represent 0.06 0.10 0.11 0.10 0.24 1.00 Solve a problem 0.09 0.22 0.44 0.07 0.39 0.02 1.00 Network with peers 0.12 0.07 0.06 0.13 0.19 0.14 0.07 1.00 Obtain technical or product 0.33 0.03 0.11 0.05 0.30 0.03 0.21 0.17 1.00 specifications Variable Factor Loadings Factor 1 Buying Factor 2 Nonbuying
b. The exploratory factor analysis based on the tetrachoric correlation matrix of attendees objectives See new products/developments 0.07 0.30 Fact finding for future purchases 0.41 0.20 Make a purchase 0.73 0.01 Attend seminars/association meetings 0.02 0.21 To see specific companie(s)/product(s) 0.40 0.68 To find products to represent 0.12 0.49 Solve a problem 0.84 0.15 Network with peers 0.20 0.73 Obtain technical or product specifications 0.11 0.33 Eigen value 2.02 1.47 Cumulative variance explained 22.40% 38.79 Note: The Eigen values of the factor analysis are 2.02, 1.47, 1.17, 1.10, 0.94, 0.85, 0.73, 0.47, 0.25. The minimum Eigen value criterion suggests a four-factor pattern but we retained a two-factor solution for better exposition.
Similarly, we conducted an explanatory factor analysis for attendees objectives. We use the tetrachoric correlation matrix as the input for the factor analysis and report the results in Table 4b. The factor pattern for attendees objectives is not as clear as that for exhibitors objectives. The minimum eigenvalue criterion would suggest a four-factor structure of our data. However, the third and fourth largest eigenvalues are 1.17 and
412
1.10, very close to the traditional cut-off threshold of the minimum eigenvalue criterion. For ease of interpretation we retained a two-factor structure. Consistent with previous research (e.g., Kerin and Cron 1987), we find that these two factors are best described by Buying factor and Non-Buying factor. Objectives such as Fact Finding for Future Purchases, Make a Purchase, and Solve a Problem are more heavily loaded on the buying factor while objectives such as See New Products/Developments, Attend Seminars/Association Meetings, Network with Peers, and Obtain Technical or Product Specifications are more heavily loaded on the Non-Buying factor. Other objectives split across factors. The two-factor model explains 38.79% of the total variance.
In the exhibitors survey, respondents reported what product types they exhibit. In the attendees survey, respondents reported what role they play in the purchase of each product type, using the same product classification. We define breadth of product interests for an exhibitor or attendee as the number of product types in which the exhibitor or attendee shows interest. Figure 2 reports the distributions of breadth of product interests for both exhibitors and attendees. We find that the means (1.23 for exhibitors
413
and 2.20 for attendees) of the breadth of product interests for both exhibitors and attendees are significantly greater than 1 (at p = 0.01), indicating significant cross product interest for both exhibitors and attendees.
The first column of Table 5 describes the results from running this model, with the observations being weighted by the total number of shows in the industry. As can be seen from the table, all variables are significant at 5% and beyond with the signs consistent with Propositions 14. The negative coefficients of EXSELL and ATBUY show that the lower the selling/buying propensity from exhibitors or attendees the higher the likelihood of horizontal shows to be formed, supporting Proposition 1. The positive coefficients of EXPROD and ATPROD show that the higher the breadth of product interests from exhibitors or attendees, the more likely horizontal shows are to be formed, supporting Proposition 2. The results on the disparity of breadth of product interests between exhibitors and attendees are interesting. First, the negative coefficient of DIFFPROD on horizontal shows in equation (2) suggests that the greater the difference in the breadth of product interest between exhibitors and attendees, the less likely horizontal shows are to be formed, directly supporting Proposition 3. Second, assuming other things being equal,
414
TABLE 5. Results of simple logit, sequential two-stage and simultaneous equations models
Model Type Simple Logit Sequential Two-Stage Simultaneous Equations
Coefficients for proportion of horizontal shows Constant 7.166*** (1.023) EXSELL 0.586** (0.238) ATBUY 0.283** (0.143) EXPROD 0.625*** (0.212) ATPROD 1.938*** (0.272) DIFFPROD 1.256*** (0.277) TECH 0.563*** (0.175) Residual Coefficients for total number of shows Constant EXSELL ATBUY EXPROD ATPROD DIFFPROD TECH Variance-Covariance of errors SIGMA1 SIGMA2 RHO Log Likelihoods 388.29
7.561*** (1.093) 0.624** (0.250) 0.255* (0.143) 0.682*** (0.217) 1.934*** (0.275) 1.110*** (0.282) 0.617** (0.192) 0.510 (0.341) 1.593 (1.284) 0.004 (.153) 0.047 (0.121) 0.329* (0.178) 0.275* (0.148) 0.274 (0.198) 0.537*** (0.115)
6.881*** (1.652) 1.188*** (0.380) 0.558** (0.261) 1.737*** (0.342) 1.906*** (0.397) 1.718*** (0.397) 0.570** (0.252)
1.593 (1.011) 0.004 (0.113) 0.047 (0.099) 0.329** (0.144) 0.275** (0.121) 0.274* (0.161) 0.537*** (0.274) 0.485*** (0.075) 4.033*** (0.926) 0.105** (0.052) 400.08
Note: Standard errors are in parenthesis. ***, **, and *denote significance at 1%, 5%, and 10%, respectively. Reported as general fit indices; not comparable across models.
415
forming smaller vertical shows over a larger horizontal show would lead to an increase in the total number of shows organized when DIFFPROD is high. Our empirical analysis further supports this argument: the coefficient of DIFFPROD on total number of shows in equation (1) is positive and significant. Finally, the positive coefficient of TECH shows that the higher level of industry technology innovativeness, the more likely horizontal shows are to be formed, supporting Proposition 4.
Downloaded By: [Romanian Ministry Consortium] At: 19:51 2 March 2010
Robustness Checks
We took several steps to assess the robustness of these results. We calibrated standard OLS and probit models and found consistent results, with the logit model demonstrating both superior fit and predictive validity. However, a concern with the simple binary logit model is the exogenous treatment of the number of shows (n) as the weighting variable because the determinants of diversity may also affect n, which makes the number of shows endogenous and the estimates biased. To assess our Propositions more rigorously, we ran an OLS model on ln n (rather than on n as n is restricted to be nonnegative), took the residual as a control of the endogeneity of n, and used it as an additional explanatory variable in the logit model.1 The result of this two-stage model is reported in the second column of Table 5. The upper half of the column reports the logit coefficients. As before, all the determinants turn out to be significant and are consistent with our Propositions. A problem with the two-stage estimation procedure is that while such a procedure produces consistent estimates, those estimates are not efficient in general, compromising our confidence in the significance of the estimates. Estimation efficiency may not change the signs of the estimates, but does change the significances, which consequently changes the supports of our Propositions. The most rigorous way to assess our Propositions involves estimating the nonlinear simultaneous equation system jointly using an MLE
ln ni = Xi g + ui ni ~ Binomial (ni , F ( Xi b + vi ))
h
(1) (2)
where Xi are the exogenous determinants of trade show formation (i.e., Constant, EXSELL, ATBUY, EXPROD, ATPROD, DIFFPROD,
416
TECH), F (.) is the logistic cdf,2 (ui, vi ) represent error terms that are
0 s2 rs1s 2 u jointly normally distributed as i ~ Normal , 1 , 2 vi s2 rs1s 2 0 and ni and nih represent the total number of shows and the number of
horizontal shows in industry i. This approach,3 although methodologically more complex, provides a more rigorous test of our Propositions. The results from this simultaneous two-equation estimation approach are described in the third column of Table 5. These results are consistent with those we obtained from the sequential two-stage estimation. The coefficients determining the log of n (reported in the middle of the column) coincide with those obtained in the two-stage model4. The correlation coefficient (reported along with the standard deviations of the error terms at the bottom part of the column) is negative and also significant at 5%. More importantly, the beta coefficients (reported on the top part of the column) are significant at 5% and beyond, and appear with signs that support our Propositions. In summary, all three approaches support our Propositions on the correlates of trade show diversity. Notice that the error correlation between the number of shows and the diversity of shows is significant (at 0.05), suggesting the need for a simultaneous approach to testing our Propositions. That correlation also shows that several of the drivers we have identified that affect show diversitybreadth of product interest for exhibitors and attendees and technological innovativeness of the industryappear to affect the number of shows in an industry as well.
417
reconcile; empirical explorations on the correlates of trade show diversity should help shed light on understanding those results in particular and on the trade show generation process in general.
Managerial Implications
One implication of our research is that it provides a framework to predict trade show diversity based upon an understanding of the level and the congruence between exhibitor and attendee objectives and breadth of product interests. To illustrate the managerial implications of that framework, we use the simultaneous equation model with a jackknife approach: for each industry we used the data from the other 20 industries as the calibration sample and predicted the proportion of horizontal shows for the single hold-out industry. The residual is the difference between the empirical proportion of horizontal shows in that industry and the predicted proportion, based on the estimated coefficients of the model from the calibration sample. Table 6 gives the results (along with full samplefitted values and residuals). From a managerial perspective, it appears that the packaging and the building and construction industries may be overserved by horizontal shows (large positive residuals) while the photographic industry appears to be underserved by those types of shows (large negative residuals).5 Thus there may be opportunities for organizers in these industries to profitably redefine the focus of some of the shows there. Hence our framework, which links trade show participants objectives and industry characteristics to the diversity of trade shows, has clear managerial uses: it can help trade show organizers diagnose and improve the appropriate mix of shows in a given industry and help trade show participants select better shows to attend, depending on their show objectives and buying/selling interests.
418
Actual Horizontal Show Proportion Estimated Horizontal Show Proportion 17.00% 1.00% 97.00% 4.00% 0.00% 95.00% 12.00% 0.00% 0.00% 94.00% 53.00% 100.00% 3.00% 100.00% 40.00% 100.00% 71.00% 0.00% 0.00% 100.00% 0.00% 23.62% 7.87% 99.48% 15.05% 12.08% 84.31% 20.46% 9.30% 9.67% 86.17% 62.40% 89.40% 13.98% 95.76% 32.42% 95.70% 74.01% 17.95% 14.70% 65.86% 10.06% 6.55% 6.94% 2.60% 10.80% 12.08% 11.00% 8.60% 9.30% 9.67% 7.45% 8.91% 10.60% 10.95% 4.24% 7.58% 4.30% 2.58% 17.95% 14.70% 34.14% 10.06% 10.05% 23.71% 7.84% 100.00% 19.29% 12.32% 82.98% 16.77% 9.63% 10.47% 87.29% 59.12% 89.41% 15.61% 95.74% 5.06% 95.68% 82.08% 18.57% 15.45% 64.72% 10.35% Residual Estimated Horizontal Show Proportion Full Sample Jackknife
Industry
Residual
Computers and computer applications Communications Engineering Medical and health care Housing Food processing and distribution Electrical and electronics Paint Nursing Automotive and trucking Chemicals Plastics Radio, TV, and cable Energy Building and construction Restaurants and food services Manufacturing Photographic Welding Packaging Education Unweighted MAD
6.64% 6.91% 3.12% 15.03% 12.32% 12.34% 4.91% 9.63% 10.47% 6.33% 5.63% 10.59% 12.58% 4.26% 34.94% 4.32% 10.66% 18.57% 15.45% 35.28% 10.35% 11.92%
419
specific correlates: breadth of product interestby attendees and exhibitors and an industry-specific correlateand the technological innovativeness of an industry. A more complete theory of trade show formation will require the identification of economic and other drivers beyond those identified here, such as other industry-specific correlates including industry competitiveness, concentration, product profitability, growth rate, and product life-cycle stage. Trade shows remain an important and understudied element of the communications mix, demanding more research to better understand how they can best be used and how to better assess their effectiveness. We hope research such as this and the research directions noted above will add to our understanding and improve our use of this important marketing communications vehicle.
NOTES
1. We also tried using a count data model to explain n instead of running an OLS model (with ln n as the left hand side variable). Both Poisson and negative binomial regression models score much worse in terms of log likelihoods (the values are 113.98 and 94.94 as opposed to 14.58 for the OLS model) and hence, AIC. Because of this finding and also because of the difficulty of extending such a model to the two-stage approach outlined next, we followed the simple OLS approach. 2. We used the cdf for the normal distribution in estimation and prediction as well and found that the cdf of the logistic distribution has better goodness-of-fit statistics and performs better on tests of holdout predictive validity. 3. See the Appendix for the model estimation procedure. 4. This is not a surprising result. OLS coefficient estimates and MLE estimates are identical for linear regression models with normally distributed errors. A close look at equation (3) in the Appendix reveals that the stage one likelihood value dominates the joint likelihood value, ensuring close coefficient estimates for stage-one for both the sequential 2-stage model and the simultaneous model. 5. From Table 6 we note that, from a predictive validity perspective, our model does quite well. More than half our sample comes from the first two categories, yet our prediction error for computers is 6.64% while it is 6.91% for communications. These compare to an average mean absolute deviation (MAD) of 11.92% for the sample as a whole (on an unweighted basis). Apart from one industry, building and construction, the full sample and jackknife predictions correspond quite well, which makes that industry a possible candidate for an outlier. (Indeed, we reran the simultaneous equation analysis after deleting this industry and found a substantial effect on our assessment of our Propositions). The outcome of the jackknife procedure demonstrates the robustness of these results, as those industries that feature disproportionately large numbers of shows do not determine the results obtained.
420
REFERENCES
Blythe, J. (2000). Objectives and measures at UK trade exhibitions. Journal of Marketing Management 16:203222. Bock, D. and M. Lieberman (1970). Fitting a response model for n dichotomously scored items. Psychometrika 135:179197. Borghini, S. F. Golfetto, and D. Rinallo (2006). Ongoing search among industrial buyers. Journal of Business Research 59(10/11):11511159. Calton, J. (1994). Comdex complaint as big computer show opens, key exhibitors grumble about costs. Wall Street Journal, November 14, A1. CEIR (2001). CEIR direct: The newsletter of the Center for Exhibition Industry Research (JanuaryFebruary). Christofersson, A. (1975). Factor analysis of dichotomized variables. Psychometrika 40:532. Dekimpe, G. M., P. Francois, S. Gopalakrishna, G. Lilien, and C. Van den Bulte (1997). Generalizing about trade show effectiveness: A cross-national comparison. Journal of Marketing 61(October):5564. Donberg, D. (2006). Plastics show visitors increase. Paper, Film and Foil Converter 80(8):18. Godar, S. H. and P. J. OConnor (2001). Same time next yearBuyer trade show motives. Industrial Marketing Management 30(1):7786. Gopalakrishna, S. and J. D. Williams (1992). Planning and performance assessment of industrial trade shows: An exploratory study. International Journal of Research in Marketing 9:207224. Gopalakrishna, S. and G. L. Lilien (1994). A dynamic model of business trade show effectiveness. ISBM Report: 31994. Gopalakrishna, S. and G. L. Lilien (1995). A three stage model of industrial trade show performance. Marketing Science 14:2242. Gopalakrishna, S. G. Lilien, J. Williams, and I. K. Sequeira (1995). Do trade shows pay-off? Journal of Marketing 59:7583. Hansen, K. (1996). The dual motives of participants at international trade shows: An empirical investigation of exhibitors and visitors with selling motives. International Marketing Review 13(2): 3953. Hansen, K. (1999). Trade show performance: A conceptual framework and its implications for future research. Academy of Marketing Science Review 99(08). Available at: http:// www.amsreview.org/amsrev/theory/hansen0899.html. The Industrial Robot (2006). Worlds largest robots and vision show increases attendance by 55 percent. The Industrial Robot 33(2):140141. Johnson, J. (2005). Trade shows stay separate. Waste News, December 19. Kerin, R. A. and W. L. Cron (1987). Assessing trade show functions and performance: An exploratory study. Journal of Marketing 51:8794. Marken, G. A. (2004). Trade show PR . . . Its time to focus on goals, substance, results. Public Relations Quarterly 49(3):3033. Muthen, B. (1978). Contributions to factor analysis of dichotomous variables. Psychometrika 43:551560.
421
Pinar, M., J. D. Rogers, and D. Baack (2002). An examination of trade show participation in a developing country: An exploratory study in Turkey. Journal of Euromarketing 11(3):3352. Sharland, A. and P. Balogh (1996). The value of non-selling activities at international trade shows. Industrial Marketing Management 25:4956. Shoham, A. (1999). Performance in trade shows and exhibitions: A synthesis and directions for future research. Journal of Global Marketing 12(3):4157. Show Daily of Conexpo and Con/Agg, 96 (1996). Program of the Conexpo and Con/Agg 96 trade show. Trade Show Bureau (1994). A guide to the US exposition industry. Denver, CO: Trade Show Bureau Resource Center. Williams, J. D., S. Gopalakrishna, and J. M. Cox (1993). Trade show guidelines for smaller firms. Industrial Marketing Management 22:265275. Zook, C. and J. Allen (2001). Profit from the core: Growth strategy in an era of turbulence. Boston: Harvard Business School Press.
422
LL(b, g, ) = Contant +
+ ln
21
i =1
1 ln ni Xi g ln F s1 s1
nh
( F ( Xi b + v)) i
(1 F ( Xi b + v))
v ni
v r s2 (ln n X g ) i i 1 s dv 1 F 2 2 s2 1r s2 1r
(3)
where f denotes the standard normal density function. Because the integral above does not admit a closed-form solution, we calculate it using numerical integration (via Gauss quadrature). Our estimates are obtained by maximizing equation (3) using Newton-Raphson algorithm.
423
424
Trade show participants (exhibitors and attendees) can use these results to determine, given their goals (breadth of product interest and buying or selling intensity), which are the most appropriate shows to attend. And show organizers, as noted previously, can use our results to identify industries that are underserved by one show type or another and uncover potentially lucrative business opportunities Although the nature of our study is exploratory, our framework offers an initial step in building up a theory of trade show formation and diversity that will ultimately benefit all participants in the exhibition industry.
Downloaded By: [Romanian Ministry Consortium] At: 19:51 2 March 2010