Market Pioneer and Early Follower Survival Risks: A Contingency Analysis of Really New versus Incrementally New Product-Markets Sungwook Min, Manohar U. Kalwani, and William T. Robinson* June 25 2005
Sungwook Min is Assistant Professor in the College of Business Administration, California State University, Long Beach, CA 90840. Tel. (562) 985-7129, fax (562) 985-5543 and e-mail:
[email protected]. Manohar U. Kalwani is the American United Life Professor of Management in the Krannert Graduate School of Management, Purdue University, West Lafayette, IN 47907. Tel. (765) 494-4400, fax (765) 494-9658, and e-mail:
[email protected]. William T. Robinson is Associate Professor of Management in the Krannert Graduate School of Management, Purdue University, West Lafayette, IN 47907. Tel. (765) 494-4432, fax (765) 496-7434, and e-mail:
[email protected].
1
Market Pioneer and Early Follower Survival Risks: A Contingency Analysis of Really New versus Incrementally New Product-Markets Abstract Does the first entrant in a new market have a difficult time surviving or do first-mover advantages provide protection from outright failure? Our empirical study of 264 new industrial product-markets yields insights into this controversial research topic. The key data analysis insights arise by comparing survival risks in markets that were started with a really new product versus an incremental innovation. When the pioneer starts a new market with a really new product, it can be a major challenge just to survive. In contrast, in markets started by an incremental innovation, market pioneer survival risks are much lower.
Interestingly, early
followers have the same survival risk across both types of markets. Overall, these results indicate that in markets started by a really new product, the first to market is often the first to fail. In contrast, in markets started by an incremental innovation, it appears that first-mover advantages protect the pioneer from outright failure.
2
Introduction Most literature reviews on survivors conclude that market pioneers tend to have a higher market share than later entrants. For examples, see Szymanski, Troy, and Bharadwaj (1995), Kalyanaram, Robinson, and Urban (1996), and VanderWerf and Mahon (1997). The typical relationship though between order of market entry and survival is controversial. It is important to try and resolve this controversy because market pioneer survival problems can easily offset their market share advantages. If so, firms can be discouraged from investing in the costly and risky attempt to pioneer new markets. Some research concludes that market pioneers experience a higher survival risk than later entrants. Olleros (1986) cites examples of unsuccessful pioneers of new technologies and observes, “again and again we see industries emerge over the dead bodies of their early pioneers” (p. 8). Lambkin and Day (1989) and Tellis and Golder (1996) predict higher failure rates for market pioneers. Lilien and Yoon (1990) and Golder and Tellis (1993) provide supporting empirical evidence. Golder and Tellis (1993), for example, report a lifetime market pioneer survival rate of only 53%. Other research indicates that first-mover advantages help protect the market pioneer from outright failure. These studies include Agarwal’s (1997) product life cycle analysis, Shepherd’s (1999) venture capitalist survey, and Robinson and Min’s (2002) analysis of industrial goods industries. For example, Robinson and Min (2002) report a ten-year survival rate of 66% for market pioneers versus 48% for early followers. Because of these divergent results, it seems likely that under some circumstances, market pioneers have higher survival rates. Under other circumstances though, they should have lower survival rates. If so, a contingency analysis may yield important insights into the complex
3 relationship between order of market entry and survival. This complex relationship is also examined by Srinivasan, Lilien, and Rangaswamy (2004). They report that network externalities have a negative impact on market pioneer survival. While network externalities are important, our contingency analysis is based on product innovation newness.
Product innovation newness is an important aspect of new product
development that has received considerable research attention. For example, see Garcia and Calantone’s (2002) literature review on innovation types. Our contingency analysis compares survival rates for markets that were started with really new products versus incremental innovations. This is part of a continuing effort to shed light on different challenges which marketing managers face in really new versus incrementally new product-markets (e.g., Urban, Weinberg, and Hauser 1996; Song and Montoya-Weiss 1998; Schmidt and Calantone 1998). When a pioneer starts a new market with a really new product, the costs and risks are unusually high, which makes survival more difficult. Because early followers can learn from the pioneer’s mistakes, they are not as vulnerable to exit. In contrast, an incrementally new product typically has lower costs and risks, which makes it easier for the market pioneer to survive. In addition, a pioneer’s temporary monopoly plus their first-mover advantages should yield higher survival rates versus early followers. Our contingency analysis is tested using data from two sources. The Thomas Register of American Manufacturers is used to identify 264 new product-markets for industrial goods and to track market pioneer (first entrant) and early follower survival over time. It is supplemented by an expert survey that estimates product innovation newness for each of these 264 new markets. The expert survey classifies 66 markets as starting with a really new product and 198 as starting with an incremental innovation.
4 In the data analysis, consistent results surface between the descriptive statistics and a hazard rate model that accounts for both observed and unobserved heterogeneity. The results reveal that market pioneers who launched a really new product have much lower survival rates than pioneers who launched an incremental innovation.
For example, the 12-year pioneer
survival rate for really new products is only 23% versus 61% for incremental innovations. One surprising result is that early followers do not have meaningful survival risk differences in really new versus incrementally new product-markets. For example, the 12-year survival rates in really new versus incrementally new product-markets are essentially equal at 38% versus 39%. This indicates that in really new product-markets, early follower learning from the market pioneer’s experience can materially reduce their survival risk. The research controversy described above can be resolved by examining survival risk differences within really new and incrementally new product-markets. In incremental innovation product-markets, the market pioneer has a consistently higher survival rate versus early followers. An opposite result occurs in really new product-markets, where the market pioneer has a lower survival rate. Overall, the key research conclusion is that the market pioneer is more likely to be the first to fail when it starts a new market with a really new product, but not when it starts the new market with an incremental innovation.
Survival Risk Hypotheses The hypotheses below highlight survival risks for market pioneers and early followers. The first two hypotheses compare survival risks across markets that were started by the pioneer’s really new or incrementally new product. The third and fourth hypotheses address survival risk within each type of market by comparing market pioneers to early followers.
5 Really New versus Incrementally New Product-Markets To assess product innovativeness, we follow Urban, Weinberg, and Hauser’s (1996) dichotomy of really new versus incrementally new products. They state, “Really new products shift market structures, represent new technologies, require consumer learning, and induce behavior changes” (p. 47). At the other end of the spectrum, an incremental innovation is designed to satisfy a felt market need and uses an existing technology or refinement of it. Really new products such as automobiles and personal computers utilize new technologies to elicit new market needs. Agarwal and Bayus (2002) point out that really new products are often unreliable and costly when they first appear on the market. Because of high levels of market and technological uncertainties, they face slow acceptance by potential users. A key source of market uncertainty arises from customers’ concern about the new product’s ability to deliver all its promised benefits.
Really new products may also require the use of
complementary hardware (e.g., roads for automobiles) or software technologies (e.g., programs for personal computers), often supplied by an independent third party, whether government or industry (Olleros 1986). Potential customers often delay adoption of a really new product until the complementary technologies become available. Really new products often entail high technological uncertainty. Potential users tend to be concerned whether a market will eventually adopt the technical standards embodied in the really new product. Also, when a dominant product design emerges for a really new product, the dominant design can take years to emerge (Christensen, Suarez, and Utterback 1998). If so, when the pioneer is committed to another product design which customers do not want, its survival can be threatened.
In sum, the resources required to overcome high market and
6 technological uncertainties associated with the adoption of really new products can cause pioneer burnout and eventual exit from the market. Incrementally new products, on the other hand, involve much lower levels of market and technological uncertainties. Market uncertainty is low since they provide incremental benefits versus existing products. Because the felt need already exists, market research can provide more accurate sales forecasts than for really new products. Technological uncertainty is limited since they tend to be based on a refinement or extension of established technologies. Thus, market pioneers who launch an incrementally new product should face low survival risk. There is some empirical evidence of high survival risk for really new products. Olleros (1986) cites pneumatic tires, sewing machines, automobiles, typewriters, helicopters, televisions, jet engines, and transistors as notable examples of product-markets in which pioneers did not survive. These examples appear to be really new products that serve unmet customer needs. The 36 product categories in Golder and Tellis (1993) span innovations such as video recorders, microwave ovens, dishwashers, personal computers, and facsimile machines. They say they have “chosen to identify pioneers in distinctive product categories rather than in narrow categories” (p. 164) which suggests their sample is more likely to contain really new products. In contrast, Urban et al. (1986) who report a zero pioneer failure rate analyze a sample that contains “tightly defined” categories such as liquid laundry detergents, instant freeze dried coffee, fabric softeners, and anti-dandruff shampoo. Robinson and Min (2002) report higher five and ten-year survival rates for market pioneers versus early followers. Their sample mainly includes incremental innovations such as disposable mops, animal access doors, boat lights, and pickle making machinery. Thus, the empirical evidence that supports higher pioneer failure rates
7 emphasizes really new product markets, while lower pioneer failure rates emphasize incrementally new markets. This yields:
H1: Market pioneers have higher survival risk for really new versus incrementally new products.
Market and technological uncertainties associated with the adoption of really new products often take years to resolve. Complementary technologies are likely to be limited until product sales reach a critical mass. The cost and risk of adopting a really new product is often very high. Such breakthrough innovations tend to be relatively expensive and deficient in performance. In view of these deficiencies and uncertainties, many potential users adopt a wait-and-see attitude. The full resolution of market and technological uncertainties may take a long time. Hence, early followers should also experience higher resistance in gaining customer acceptance in really new product-markets. This yields:
H2: Early followers have higher survival risk for really new versus incrementally new products.
Order of Market Entry Is the first to market the first to fail? Comparing market pioneer and early follower survival rates within really new and incrementally new product-markets addresses this controversial research topic. For really new products, a market pioneer faces numerous firstmover disadvantages. First-mover disadvantages in Lieberman and Montgomery (1988) span:
8 (1) free rider effects, (2) market and technological uncertainties, (3) changing technology or customer needs, and (4) incumbent inertia. Because early followers in our sample typically entered within two to three years of the pioneer, they are also vulnerable to incumbent inertia. Thus, the other first-mover disadvantages are highlighted below. Imitation is typically less expensive than innovation, so early followers have the opportunity to “free ride” on the pioneer’s investments. This should be especially true for really new products where pioneering can be costly. As mentioned above, market and technological uncertainties are often high for really new products, which can require huge resources and cause pioneer burnout (Olleros 1986). Furthermore, Kerin, Varadarajan, and Peterson (1992) suggest “the greater the [market] uncertainty level, the lower the likelihood that a first-mover will make sizable investments in capacity to achieve scale-dependent cost advantages (p. 43).” Finally, changing technology and customer needs can provide early followers an opportunity to leapfrog the pioneer with a superior product offering. The product superiority can arise from a technologically superior product or from a superior job of satisfying unmet customer needs. Two industry studies provide empirical evidence for really new product-markets. Across five diagnostic imaging product categories, Mitchell (1991) finds that later entrants have lower survival risks. In the rigid disk industry, Christensen, Suarez, and Utterback (1998) also report that later entrants have lower survival risks. Based on this theoretical and empirical evidence, we predict:
H3: For really new products, market pioneers have a higher survival risk versus early followers.
9 While the disadvantages for market pioneering are especially strong for really new products, market pioneering advantages can also be especially strong. First, market pioneers benefit from a temporary monopoly, which provides temporary protection from competition. A temporary monopoly also helps pioneers strengthen their first-mover advantages (Brown and Lattin 1994; Huff and Robinson 1994). Because imitation should be more difficult for really new versus incrementally new products, this tends to lengthen the pioneer’s monopoly. Second, first-mover advantages help protect the pioneer from outright failure. Pioneer first-mover advantages span numerous sources. These sources include a stronger brand name (Schmalensee 1982), customer preferences that are shaped to favor the pioneer (Carpenter and Nakamoto 1989), and broader product lines (Robinson and Fornell 1985). In really new product markets, it seems likely that these first-mover advantages will tend to be stronger. For example, a really new product is more unique versus an incremental innovation. This uniqueness should tend to strengthen consumer memory and brand name recall. Unique products should be more effective in shaping customer preferences in favor of the pioneer. Finally, unique products should have more opportunities for product line extensions. Overall, when longer pioneer leadtimes and stronger first-mover advantages outweigh first-mover disadvantages, we have:
H3a: For really new products, market pioneers have a lower survival risk versus early followers.
For incrementally new products, these first-mover disadvantages should be modest. First, there should be fewer opportunities to “free ride” on the pioneer’s investments. This is because
10 these investments are typically modest in comparison to developing a really new product. Second, incremental innovations typically refine and modify existing products, so the felt need for these products often exists and the technologies are well established. With established customer needs and technologies, market and technological uncertainties are modest. Finally, changes in customer needs and technologies are less likely to arise and to surprise the pioneer. In addition, market pioneers should still benefit from a temporary monopoly and firstmover advantages. Even an incremental innovation cannot typically be copied overnight, so the temporary monopoly makes it easier to survive. Also, first-mover advantages in terms of brand name strength, customer preferences that are shaped to favor the pioneer, and broader product lines should make it easier to survive. Empirically, as discussed above, Robinson and Min (2002) and Urban et al. (1986) report relatively high market pioneer survival rates. Because both studies have relatively narrow market boundaries, they should be dominated by incremental innovations. This yields:
H4: For incremental innovations, market pioneers have a lower survival risk versus early followers.
While the pioneering disadvantages discussed above are not as great in incrementally new product markets, Golder and Tellis (1993) describe three additional disadvantages. First, improper positioning arises when the pioneer misses the ideal point and repositioning costs are high. Even for incrementally new product-markets, later entrants have a better chance of hitting the ideal point. Second, resource requirements can change over time. An early follower is in a better position to predict the key resource requirements for survival. Finally, a market pioneer
11 may not be able or willing to commit the resources necessary to succeed. Again, an early follower is in a better position to understand the resources required for success. Thus, when the pioneering disadvantages outweigh the advantages discussed above, we have:
H4a: For incremental innovations, market pioneers have a higher survival risk versus early followers.
Data The Thomas Register data have been used in numerous survival studies. Examples in the economics literature include Gort and Klepper (1982), Agarwal (1997), and Agarwal and Gort (1996, 2001). In the marketing literature, it has been used by Agarwal and Bayus (2002) and Robinson and Min (2002). The Thomas Register is a national buying guide that provides “a comprehensive detailed guide to the full range of products manufactured in the United States” (Lavin 1992, p. 129). The Thomas Register attempts to achieve comprehensive coverage by subscribing to a broad range of industry newsletters, searching for start-up ventures in university incubators, and by providing a free listing in each annual issue. For instance, the 2003 Thomas Register includes approximately 173,000 firms and more than 65,000 product categories. A key Thomas Register data strength is that the market boundaries are buyer-driven. To help industrial buyers, a new product category is typically added to a group of closely related categories. A new category does not split an existing category. For example, an industrial buyer searching for valves will find air valves, nuclear valves, frost-proof valves, solar collector draindown valves, etc.
12 The Thomas Register’s professional staff determines market boundaries and the relevant competitors in each market. The Thomas Register market boundaries are narrower than Golder and Tellis (1983), who include broad categories like microwave ovens, cereal, and soft drinks. They are typically broader than Urban et al. (1986), who examine consumer packaged goods such as light beer, antacids, and steak sauce. Because it is a national buying guide, the Thomas Register typically excludes firms with exclusively local sales.
By excluding local markets, our sample emphasizes regional and
national markets. In terms of scale of entry, this operational definition falls between Urban et al. (1986), who require national distribution and Golder and Tellis (1993) who include local markets. New markets are identified by comparing annual editions of the Thomas Register from 1960 to 1995. A new market is identified when it is listed in one year, but not in prior years. Identifying 264 new industrial markets was the most time consuming part of the data collection process.
Thomas Register Data Limitations While it is unlikely that the Thomas Register staff will set market boundaries that are too broad, what happens when the market boundaries are too narrow? In these markets, our sample misses the market pioneer who has already entered a broader product-market. Thus, our research inadvertently compares an early follower versus a later entrant or even a later entrant versus other later entrants. In this situation, the true results should be even stronger than those reported below. This is because the importance of order of entry should have diminishing marginal
13 returns. For example, greater differences should arise between the market pioneer and an early follower than between two later entrants. A second data limitation arises from our definition of exit. We follow Agarwal’s (1997, p. 574) guidelines with an exit arising when the firm’s name and address are both missing from the product category in a given year. This avoids the problem of confusing an exit with a name change, such as a conglomerate merger, or confusing an exit with an address change from office relocation. Even so, when a successful firm is merged or acquired, its name and address are often changed. In this case, it is classified as an exit. Because exit implies failure rather than success, this limitation surfaces in the other Thomas Register survival studies cited above. In an attempt to quantify this potential bias, we gathered additional data for the most recent Thomas Register exits that arose between 1994 and 2002. There were 178 product exits during this nine year period (see Table 1). Cases were then identified when a firm exit arose in the same year as the product exit. A firm exit arises when a firm is de-listed from the Thomas Register. This approach assumes that when a merger or acquisition occurs, a product exit and the firm exit arise in the same year. In Table 1, only 25% of the product exits were accompanied by a firm exit. Also, there are no meaningful differences associated with either order of entry or product innovation newness. Some additional insights are provided for the 45 cases when a product exit and the firm exit arise in the same year. These insights are based on a Google search, an Internet search for bankruptcy court filings, and by calling the most recent telephone number listed in the Thomas Register.
14 For these 45 cases, 10 were the result of an either a merger or an acquisition, 4 were bankruptcies, and 31 had an unknown outcome.1
For the mergers and acquisitions, it is
important to know if the acquired firm was successful or unsuccessful in the product category that was being tracked. Successful product category performance should not be classified as an exit, whereas unsuccessful performance can be classified as an exit. While there is not enough data to determine success at the time of exit, we assume that when product category performance is successful, the acquiring firm should remain active in the product category through 2002. (This is roughly three years after the acquired firm’s exit.) In Table 1, the acquiring firm remained active in the product category for only 2 out of the 10 mergers and acquisitions. This suggests that even when a Thomas Register exit arises from a merger or an acquisition, the acquired firm is more likely to be struggling than successful. Overall, these results indicate that mergers and acquisitions of successful firms do not seriously bias the Thomas Register exit measure.
Key Definitions In each product-market, the market pioneer is defined as the market’s first entrant. Since our research compares market pioneers to early followers, our analysis excludes markets with multiple first year entrants. This is because a unique first entrant (market pioneer) cannot be identified. When a unique first entrant is identified, the market is tracked forward on an annual basis until another entrant is identified. This entrant and any other entrants in that year are called early followers.
1
For the 31 cases with unknown outcomes, the acquired firm could not be contacted by calling the most recent telephone number listed in the Thomas Register. Because successful firms should attempt to maintain telephone contact with established customers, this outcome seems more likely to arise for unsuccessful firms.
15 Industry and research experts were surveyed to estimate the degree to which each new product-market was started with a really new product or an incremental innovation. Urban, Weinberg, and Hauser’s (1996) conceptual distinctions were read and understood. In these conceptual distinctions, a really new product creates a new market, relies on new technology, and requires customer learning. An incrementally new product is designed to satisfy a felt market need and uses an existing technology or refinement of it. In the rating scale designed to elicit the newness of a product innovation, 1 is an incremental innovation and 7 is a really new product. Table 2 provides the survey questionnaire for chemicals and allied products. For each of the 264 new markets, four to nine experts rated the degree of product innovativeness when the new market started. Fifteen of the 39 experts had a Ph.D., most had industry work experience, and all were pursuing scientific research.
Roughly 90% were
employed in a research oriented university which has most of its engineering departments ranked in the top 20 nationally. Each expert only evaluated products in their field of expertise.2 To the best of our knowledge, the experts gave ratings that were internally consistent across the product categories. Even so, in the questionnaire pretest, some experts in basic research labs gave consistently low ratings for the degree of product innovativeness. To control for this bias, our final survey included some well-known really new products that serve as benchmarks.
These benchmarks were mainly derived from earlier research studies.
For
chemical and allied products, for example, our survey included DDT, saccharin, penicillin, and streptomycin (Gort and Klepper 1982). In other categories, we included the personal computer,
2
The surveys were done in person. While it was not part of the formal survey, in most cases, the expert could not identify the market pioneer or the first early follower. This may arise because the average year of market inception was in the 1970s. Thus, it does not appear that early entrant survival knowledge biased the degree of product innovativeness measure.
16 laser jet printer, Walkman, Web TV (Urban, Weinberg, and Hauser 1996), transistors, electric shavers, and radar detectors (Gort and Klepper 1982). To adjust for individual response bias, the first step computes each respondent’s benchmark score, which is the average score across the well-known really new products. The second step subtracts the respondent’s benchmark score from the corresponding raw score. For each expert, a product is considered really new when the raw score is greater than or equal to the benchmark score. A product is classified as really new when the majority of experts agree that a given product is really new. This procedure yields 66 really new products and 198 incremental innovations. To quantify the reliability of the product innovation newness measure, Cronbach’s α estimates the average correlation in inter-expert ratings. Table 3 reports the results across twodigit SIC code industries. Cronbach’s α ranges from 0.76 to 0.96, which indicates an acceptable degree of reliability. Overall, the benchmark adjustment to the product innovativeness measure appears to reduce individual response bias. This is because it yields a reliable measure and tends to strengthen the hazard rate modeling results below.
Descriptive Statistics Across the entire sample, the market pioneer 12-year survival rate is 52%. Golder and Tellis (1993) report a lifetime market pioneer survival rate of 53%. Does this imply the “first to market is the first to fail?” It turns out, however, that in our entire sample, the 12-year early follower survival rate is only 39%. Because market pioneers have significantly higher survival rates, it appears that first-mover advantages help protect the pioneer from outright failure. While
17 this conclusion applies to the entire sample, more accurate insights arise by comparing really new product-markets versus incremental innovations. In our sample, the really new product-markets have 66 market pioneers and 146 early followers. The incremental innovation product-markets have 198 market pioneers and 340 early followers. The average year of market inception is 1978 for really new products and 1975 for incremental innovations.
Average pioneer leadtime, which is the length of the pioneer’s
monopoly, is only 2.5 years. Table 4 describes the sample in terms of two-digit SIC codes and new product newness. The really new products include thirteen chemical products such as gelling agents, cellulose propionate, indium compounds, dielectric gas, parylene, etc. Really new products in other industries include voice activated microphone, osmometers, blood pressure monitors, video microscopes, bomb detectors, and computer screen projectors. The incremental product-markets include welding nozzles, poultry chillers, asphalt concrete, boat coolers, bottle sorters, fiberglass washers, boat loaders, and ice dicing machinery. The appendix provides a complete list of the product-markets, the year of market inception, and the degree of product innovativeness. Using 4, 8, and 12-year survival rates, Table 5 provides initial hypothesis testing insights. Longer time horizons are not used because right-censoring problems limit the sample size. Twelve-year survival rates provide the most information, so they are highlighted below. For H1, 12-year survival rates for market pioneers are much lower for really new products than incremental innovations. At 23% versus 61%, incremental innovation survival rates are almost three times higher versus really new products. This difference can be explained by the greater market and technological uncertainties that pioneers have to overcome when they launch a really new product. This result provides strong support for H1.
18 For H2, it is surprising that there are no meaningful survival rate differences for early followers in really new versus incrementally new product-markets. This result is also surprising because these early followers typically entered the new product-markets only two to three years after the pioneer. This suggests that early followers benefit from the efforts and experience of the pioneer in overcoming market and technological uncertainties. For H3, market pioneers in really new product markets have significantly lower 12-year survival rates than early followers, 23% versus 38%. For H4, an opposite sign arises with market pioneers having significantly higher 12-year survival rates (61% versus 39%). The higher levels of market and technological uncertainties in really new product markets, which should decrease market pioneer survival rates, can explain this sign reversal. Two industry examples help illustrate these results. A voice activated microphone was classified in our expert survey as a really new product. It was pioneered by Sherwood Communications Associated Ltd. in 1992. It entered the market as a startup business with roughly one million dollars in assets. It exited the market in 1993, which is the same year the first early follower, Telephonics Corporation., entered. Telephonics was an industry incumbent, which had experience in manufacturing headsets and microphones. Its asset size was over $250 million. Telephonics survived through 2002 and was one of active players in the market. A second example is a welding nozzle, which in our expert survey classified as an incrementally new product. It was pioneered by Maryland Ceramic and Steatite Co. in 1987, which was an industry incumbent with over one million dollars in assets. In 2002, it was one of 22 market survivors. The first early follower was Lurmark Ltd., a start-up business. It entered the market in 1990, but exited in 1993.
19
Hazard Rate Analysis While descriptive statistics are useful, a multivariate hazard rate analysis provides more accurate hypothesis testing results. A hazard rate is the conditional likelihood of a firm’s failure in time period t, given that the firm has survived through time period (t-1). Two widely recognized benefits of a hazard rate analysis are: (i) representing the impact of time varying explanatory variables on a firm’s likelihood of failure, and (ii) addressing the right censoring problem (Helsen and Schmittlein 1993, p. 397). Right censoring occurs when a firm does not have a completed spell within the data observation window and all that we know is that its survival duration lasted from the time of its market entry until the end of the observation window. Because our data are annual observations, we employ a discrete hazard rate model which incorporates the impact of time varying covariates (e.g., competitive firm density in a product-market) on a firm’s likelihood of failure. We model the impact of time on a firm’s likelihood of failure non-parametrically in order to capture any arbitrary pattern in the firm’s baseline hazard rate over time (see, for example, Han and Hausman 1990; Meyer 1990). In other words, we assume the baseline hazard rate function is a step function of time that is constant within a year but is allowed to vary from one year to another. An advantage of this approach is that it allows consistent estimation of the effect of explanatory variables for arbitrary underlying baseline hazard functions. Finally, the hazard rate model incorporates the impact of unobserved heterogeneity on firm hazard rates. In the hazard rate model,3 let t denote the survival duration of an entrant with the cumulative probability distribution function F(t).
The associated hazard rate and survival
functions are h(t) and S(t)=1-F(t). Now, consider firm i that we are able to observe through a 3
Vanhuele et al. (1995) use the same hazard rate model to analyze marriage dissolution data. Their paper has a more complete hazard rate model description.
20 total of ti time periods (years in our case) where ti may range from one year to the entire length of the data observation window, say, m years. That is, assume that firm i has survived through the year (ti-1) and then we observe it into the tith period, [ti-1, ti), to see if it failed during the year or was right censored during the period. If firm i is found to exit during the t ith year, the probability of its failure in the t ith period is [S(ti-1) - S(ti)]. If, on the other hand, it was censored during the
t ith year, all that we know is that the firm survived through the end of period (ti-1); the probability of that event then is S(ti-1), under the assumption that censoring takes place at the beginning of the interval. Hence, we can write the likelihood function for firm i which we are able to observe into the t ith year: (1)
L(t i ) = [S (t i − 1) − S (t i )]
1−δ i
× [S (t i − 1)] i , δ
where the indicator variable δ i takes the value one if firm i is right censored during the t ith year and zero if firm i is exits the market during the t ith year. Multiplying and dividing the right hand side of Equation (1) by [S (t i − 1)] i
δ −1
we obtain
S (t − 1) − S (ti ) L(ti ) = i S (ti − 1) (2)
= [h(t i )]
1−δ i
1−δ i
× [S (ti − 1)]
× [S (t i − 1)] ,
where h(ti) is the hazard rate function representing the chance of firm i’s exit in the t ith period given it has survived through the period ti-1. We now explain our derivation of the expression for hazard rate function, h(ti). We will then obtain an expression for the associated survival function S(ti-1).
21 Overall, the hazard rate of firm i in time period ti, will depend on the baseline hazard rate function and the firm’s covariate values in time period ti, Xi(ti). Firm i’s hazard rate in time period ti can be expressed as: (3)
h(t i ; X i (t i ) ) = λ0 × ϕ (t i ) × φ ( X i (ti ) ) ,
where λ0 is the base hazard rate, ϕ (t i ) is a step function which captures time (or age) dependence of the hazard rate, and φ ( X i (ti ) ) is a time-varying covariate function.
As
mentioned, the age dependence function ϕ (t i ) is modeled as a step function which remains constant within a year equal to, say, Ct in year t, but is allowed to vary from year to year. The time dependence component of the hazard rate function, ϕ (t i ) , in year ti is: (4)
ϕ (t i ) = exp[C t D(t i )] , i
where the dummy variable, D(ti), equals one in year ti and zero in all other years. No dummy variable is used for the first year because it is not possible to identify both λ0 and C1 Therefore,
λ0 is the time dependence component of the hazard rate function in year one. Positive (negative) estimates of the time varying Ct coefficients for the other years indicate a higher (lower) hazard rate relative to the first year hazard rate, λ0 . A firm’s hazard rate will also depend on its characteristics (e.g., its order of entry in a product-market) and product-market factors some of which may vary with time such as competitive firm density. The impact of covariates in firm i’s hazard rate in period ti is: (5)
φ ( X i (ti ) ) = exp[βX i (ti )] ,
where Xi(ti) is a vector of (possible) time-varying covariates and β is a vector of the covariate coefficients. A positive estimate of a β -coefficient implies that an increase in the value of the
22 covariate will enhance the likelihood of firm failure.
Specifically, when the jth covariate
increases by one, the conditional probability of firm failure changes by [exp(βˆ j ) − 1] × 100% . Substituting for the terms ϕ (t i ) and φ ( X i (ti ) ) from Equations (4) and (5) in Equation (3), we obtain h(t i ; X i (t i )) = λ0 exp[C t D(t i ) + β X i (t i )] .
(6)
The survivor function associated with the hazard rate in Equation (6) is given by (Lancaster 1990; Yamaguchi 1991): S (t i ) = exp[− H (t i )] ,
(7) where
ti
H (t i ) = ∫ h(u )du .
(8)
0
While our time-varying covariates (e.g., competitive firm density in a product-market) can change from year to year, we assume they remain constant within each year.
Given this
simplifying assumption, Vanhuele et al. (1995) show that Equation (8) can by rewritten as, H (t i ; Bi (t i )) = λ0 Bi (t i ) ,
(9) ti
where Bi (t i ) = ∑ exp[βX i (t ) + Ct D(t )] . Substituting from Equation (9) into Equation (8) yields, t =1
S (t i ; Bi (t i )) = exp[− λ0 Bi (t i )] .
(10)
We can now substitute for the survival function from Equation (10) into the expression for the likelihood function in Equation (1), and obtain the log likelihood function for the N firms in our sample (see Vanhuele et al. 1995), N
(11)
LL = ∑ {(1 − δ i ) log[exp(−λ0 Bi (t i − 1)) − exp(−λ0 Bi (t i ))] − δ i λ0 Bi (t i − 1)} . i =1
23 The final step in our hazard rate analysis is to account for unobserved heterogeneity. A firm’s likelihood of failure may depend on some factors that are not measurable and others for which data are not available. Thus, firms may have different mean survival rates even when their observable characteristics are the same. Unobserved heterogeneity in the base hazard rate is modeled by allowing λ0 to vary across the population according to a time invariant distribution. Thus, the likelihood function contributions, derived above are conditional on a specific value of λ0 (see Equation 11). The unconditional likelihood function is obtained by weighting the conditional likelihood by the relative occurrence of λ0 . This is specified as, ∞
L(t i ) = ∫ L(t i | λ0 ) g (λ0 )dλ0 ,
(12)
0
where g( λ0 ) is called the (unobservable) mixing distribution. The gamma distribution, because of its flexibility, is often used as the mixing distribution (Lancaster 1979; Heckman and Singer 1984). Vanhuele et al. (1995) show that assuming a gamma mixing distribution leads to a closed form solution for the log-likelihood function: (13)
αr αr LL = ∑ log (1 + δ i ) − [Bi (t i − 1) + α ]r Bi (t i − 1) + (1 − δ i ) exp(βX i (t i ) + Cti D(t i )) + α i =1 N
[
, r
]
where α and r are the scale and shape parameters of the gamma mixing distribution. The coefficient of variation of the mixing distribution is often interpreted as a measure of the degree of heterogeneity. For the gamma distribution CV reduces to
CV =
SD[λ0 ] = E[λ0 ]
r
α2 = 1 , r
α
r
24 giving the r parameter, the interpretation that a low (high) r-value suggests relatively more (less) heterogeneous population.
Control Variables
Table 6 provides the variable definitions, means, and standard deviations for the full model specification. Key model specification variables used in the hypothesis testing are the really new product-market dummy variable, a market pioneer dummy variable, and a really new product-market * market pioneer interaction term. All the control variables are discussed below. Conventional Economic Variables: Geroski’s (1995) literature review suggests that the likelihood of survival is greater for larger firms than for smaller firms. The Thomas Register includes the tangible asset size for each firm. Because the average tangible asset size varies over industries and has changed over time, entrant size is divided by its industry average. The industry is defined at the two-digit SIC code level. Incumbents are likely to have higher survival rates than start-up ventures because of their prior experience, market knowledge, brand equity, and customer relationships.
We follow
Agarwal and Gort (1996) and define incumbents as firms in existence in year t-1. In recent decades, increasing international competition and shorter new product development cycle times have led to an increase in competitive intensity (Griffin 1997, p. 450). If competitive intensity has increased from the 1960s to the 1990s, then hazard rates should be positively influenced by the calendar year of market inception. Abnormally high hazard rates in the years of economic decline are controlled with a time varying dummy variable. The dummy variable equals one in years when there is a decline in the annual GDP growth rate. Audretsch (1991) finds that the 10-year firm survival rates vary
25 considerably across different industries. Thus, the hazard rate model also includes four industry specific dummy variables. Population Ecology Variables: Population ecologists predict two contrasting effects of “density dependence” on firm hazard rates – legitimation and competition (Hannan and Freeman 1989, Hannan and Carroll 1992). Legitimation of a product innovation increases as the number of firms in the new product-market increases, thereby resulting in the decline of hazard rates of all firms. There is, however, a ceiling on the impact of the legitimacy process. At higher levels of firm density, legitimation reaches a saturation level and the competitive effect of firm density dominates, resulting in higher hazard rates for all firms. Overall, the population ecology theory predicts a U-shaped relationship between firm density and hazard rate in new product-markets. Hence, the hazard rate model includes a firm density and a firm density squared term.4 Density dependence theory also predicts a persisting competitive effect of firm density at the time of a firm’s entry into a product-market. Carroll and Hannan (1989) attribute this effect to two sources, “resource scarcity” and “tight niche packing.” They contend intense competition at the market entry creates condition of resource scarcity. With scarce resources, market entrants that find it difficult to move to full-scale operation face strong selection pressures. A second consequence of intense competition at the time of market entry, they argue, is tight niche packing. Because market entrants seldom compete head-to-head with market incumbents, they are forced to pursue marginal target-market opportunities. Carroll and Hannan (1989), Carroll et al. (1996), and Banbury and Mitchell (1995) provide empirical evidence supporting the adverse impact of firm density at the time of market entry on the probability of firm survival.
4
To adjust for differences across product markets, firm density is normalized by dividing it by the average number of competitors over the 10-year period following the end of the pioneer’s monopoly.
26 Strategic Variables: Because increasing pioneer leadtime tends to strengthen first-mover advantages, pioneer leadtime is included in the hazard rate model. Increasing pioneer leadtime should decrease a market pioneer’s hazard rate. For early followers, increasing pioneer leadtime can increase their hazard rates as pioneers gain strength with longer leadtime. However, a short delay can help them survive because in the first year or two of the market’s evolution, decreased market and technological uncertainty can yield substantial benefits. In sum, we predict a Ushaped relationship for early follower hazard rates and pioneer leadtime (Robinson and Min 2002). Pioneer leadtime’s natural logarithm is used to adjust for diminishing marginal returns. The full specification of the hazard rate’s time-varying covariate function is shown below. For ease of presentation, deleting the i subscripts in the right hand side of the expression for φ ( X i (ti ) ) simplifies Equation (5)’s covariate function. (14)
φ ( X i (ti ) ) = Exp[ β 0 + β1 (Really New Product-Market) + β 2 (Market Pioneer) + β 3 (Really New Product-Market * Market Pioneer) + β 4 (Market Pioneer * Ln Pioneer Lead Time) + β 5 (Early Follower * Ln Pioneer Lead Time) + β 6 (Early Follower * Ln Pioneer Lead Time2) + β 7 (Ln Firm Size) + β 8 (Incumbent) + β 9 (Ln Calendar Year of Market Inception) + β 10 (Economic Declinet) + θ 1 N t + θ 2 N t2 + θ 3 N f S
+ λs ∑ Industry s ] , s =1
where Nt is the normalized firm density at time t, N t2 is the square of the normalized firm density at time t, and Nf is the normalized firm density at the time of firm i’s entry into the market. In Equation (14), the really new product-market variable equals 1 for a really new product-market, 0 for an incremental innovation product-market. The market pioneer dummy variable equals 1 for the first entrant, 0 for an early follower. In Equation (14), the derivative of
27 the logit of φ ( X i (ti ) ) with respect to really new product-market is β1 + β 3 (Market Pioneer). Hence, the estimated value of β1 + β 3 tests H1, namely, the effect of really new versus incrementally new product on the hazard rate of a market pioneer. If β1 + β 3 is positive, then pioneers have a higher exit rate for really new versus incrementally new products. For H2, if β 1 is positive, then early followers have a higher hazard rate in really new versus incrementally new product-markets . To test H3, namely, the effect of order of entry on survival risk in really new productmarkets, we need to consider the differential impact of the order of entry--market pioneer versus early follower--on the hazard rate in Equation (14). Subtracting the derivative of φ ( X i (ti ) ) with respect to early follower from the derivative of φ ( X i (ti ) ) with respect to market pioneer, we find that
[(β
2
to
test
H3
we
need
to
consider
the
sign
and
significance
of
]
+ β 3 + β 4 Ln Leadtime) − ( β 5 Ln LeadTime + β 6 Ln LeadTime 2 ) . If this estimate is positive
and significant, it supports H3 and indicates that the survival risk for market pioneers versus early followers
[(β
2
is
higher
for
really
new
products.
Similarly,
)]
(
+ β 4 Ln LeadTime ) − β 5 Ln LeadTime + β 6 Ln LeadTime 2 , tests H4.
the
estimate
of
If this estimate is
negative and significant, it supports H4 and indicates that pioneers experience lower hazard rates for incremental innovations than early followers.
Results Table 7 displays the hazard rate model results. Statistical significance is based on conservative two-tailed tests. H1 predicts that market pioneers have a higher survival risk in really new product markets. At the bottom of Table 7, market pioneers experience significantly
28 higher survival risk for really new versus incrementally new products. Specifically, the survival risk of market pioneers is 189% ([exp (1.06) –1] * 100%) higher for really new versus incrementally new products. Recall there is almost a threefold difference in the 12-year pioneer survival rate for really new versus incrementally new products, 23% versus 61%. This indicates that for market pioneers, the challenge of resolving market and technological uncertainties in really new product-markets often outweighs the benefits of first-mover advantages. H2 predicts a higher survival risk for early followers in really new product-markets. In Table 7, the estimate has the expected sign, but is not statistically significant. This result is consistent with the descriptive statistics where 12-year early follower survival rates are 38% for really new versus 39% for incrementally new products. These results indicate that the typical two to three year delay in entering a really new product market helps an early follower resolve key market and technological uncertainties. For really new products, H3 predicts that market pioneers have a higher survival risk than early followers. At the bottom of Table 7, the difference is tested at the average pioneer leadtime level of 1.98 years for the really new product sample. It is statistically significant at the 10% level. This difference is also managerially significant with a 54% ([exp (.43) - 1] * 100%) higher survival risk than early followers.
This indicates that in really new product markets, the
challenge of managing the market and technological uncertainties often outweigh the benefits of first-mover advantages. In contrast, H4 reveals that for incremental innovations, market pioneers have a substantially higher chance of survival than early followers. These market pioneers experience a 36% (= [exp(−.44) − 1] * 100%) lower survival risk than early followers, which is significant at
29 the 5% level. This supports the prediction that for incrementally new products, market pioneers have a higher chance of survival. Turning now to the coefficient estimates of the control variables, they all have the expected signs. The impact of leadtime on the pioneer survival chance has the expected negative sign and is statistically significant at the 13% level. This indicates that increasing pioneering leadtime reduces a market pioneer’s survival risk. The results support a U-shape relationship between an early follower’s hazard rate and pioneer leadtime. Thus, a short delay appears to help an early follower, but an additional delay increases an early follower’s survival risk. Incumbency significantly decreases survival risk. Further, the calendar year of market inception has the expected positive sign and is statistically significant. This indicates increasing survival risk in markets started in recent years. There is also a significantly higher survival risk in years of economic decline. The estimates of θ 1 and θ 2 show the impact of firm density on a firm’s hazard rate. As expected, the estimates for θ 1 and θ 2 are negative and positive respectively. Both estimates are statistically significant at the 1% level.
This U-shape relationship indicates that an initial
increase in the number of firms increases its legitimation, thereby lowering firm hazard rates. At higher levels of firm density, legitimation reaches a saturation level and the competitive effect dominates with an associated adverse impact on firm survival rates. As predicted, the estimate of θ 3 is negative and statistically significant at the 1% level. This indicates that an increase in firm density at the time of market entry elevates a firm’s hazard rate. Population ecologists attribute this result to the liability of resource scarcity and to tight niche packing, which makes it difficult for an entrant to find profitable positioning opportunities.
30
Robustness of Results Because theory does not provide clear guidance on the functional relationship between the degree of product innovation newness and survival, two additional functional forms are considered below. The first maintains the dichotomy used above, but only considers a product as really new when every expert classifies it as really new. This differs from the measure above which is based on the majority of expert opinion. When every expert classifies a product as really new, our sample has 30 “really” really new products. The second functional form uses a continuous measure of product innovation newness. This follows research such as Chandy and Tellis (1998) and Chandy and Tellis (2000) whose measure of radical product innovations is continuous. Our continuous measure equals the difference between the expert’s rating and the expert’s average benchmark rating.
These
differences are averaged across the experts who evaluated each of the 264 new markets. Table 8 provides the hypothesis testing results for these alternative functional forms. Model I tests the dichotomy between “really” really new products and incremental innovations. (The 36 really new product markets that do not have uniform expert support are excluded from the data analysis.) Model II uses the continuous measure of new product newness. Overall, the hypothesis testing results are consistent using all three measures of product innovation newness. This is because the results consistently indicate that market pioneers have higher hazard rates in markets that were pioneered with a really new product, but lower hazard rates in markets that were pioneered with an incremental innovation. In addition, there is no material difference between early follower hazard rates in really new versus incremental innovation markets.
31
Dynamic Survival Risk Insights While the hazard rate model above yields important survival risk insights, it does not pinpoint how survival risk changes over time. Two types of dynamic insights are developed below. The empirical hazard rate equals the number of firms that failed in year t divided by the total number of firms that were at risk year t-1. Because empirical hazard rates do not partial out the impact of additional explanatory variables or correct for unobserved heterogeneity, predicted hazard rates are also provided. The predicted hazard rates are based on four hazard rate models. The four models are estimated for market pioneers (n=66) and early followers (146) in really new product-markets plus market pioneers (n=198) and early followers (n=340) in incrementally new markets. The full model specification includes Table 7’s additional explanatory variables.
Because the
incremental innovations sample is larger, annual dummy variables for the first 12 years of commercialization are used versus 10 years for the smaller really new product-markets sample. For really new product-markets, Figures 1a and 1b display the empirical and predicted hazard rate plots. These figures illustrate how hazard rates for pioneers and early followers are initially similar, but then diverge.
The divergence arises after the sixth year of
commercialization. Note the tenth year includes year 10 plus all subsequent years. This pattern can arise when it takes a few years for the market pioneer to learn if their initial investments are on target. During the wait, the market pioneer will fight to survive. But once it is clear that their initial investments are off target and cannot be salvaged, it is time to cut their losses and move on to better opportunities. For incrementally new product-markets, Figures 2a and 2b show that market pioneers have the greatest survival advantages in the first six to eight years of commercialization. This
32 can arise when market pioneers benefit from their temporary monopoly in the first two or three years of commercialization. Also, because first-mover advantages often dissipate over time (Robinson 1988), it is not surprising that hazard rate differences narrow with the passing of time. Even so, market pioneers have consistently lower hazard rates versus early followers.
Discussion When a market pioneer starts a new market with a really new product, survival is a major challenge. This conclusion is based on 66 really new product-markets, where only 23% of the market pioneers survived for at least 12 years. In contrast, survival is much easier when the pioneer starts a new market with an incremental innovation. This conclusion is based on 198 incrementally new product-markets, where 61% of the market pioneers survived for at least 12 years. These results indicate that market pioneers have to overcome greater resistance in building primary demand for really new products. The market pioneer’s temporary monopoly, on the other hand, can help the pioneer strengthen its first-mover advantages for incrementally new products that represent extensions of existing technologies and generally do not require consumer learning. In really new product-markets, is this high degree of uncertainty also fatal for early followers? In our sample, most of the early followers trailed the market pioneer by three years or less. It is surprising that even with this relatively short delay, there is not a meaningful survival rate difference across really new and incrementally new product-markets. For example, 12-year early follower survival rates in really new product-markets are 38% versus 39% in incrementally new product-markets. This suggests that in really new product-markets, early followers can
33 benefit from the prior efforts of the pioneer in building primary demand. They can also benefit from the pioneer’s experience in addressing customer needs and concerns which can become public knowledge through professional and trade journals (Gort and Klepper 1982). Finally, early followers are more likely to leapfrog the pioneer in really new product-markets that often undergo significant product improvements during the emergent stages (Christensen, Suarez, and Utterback 1998). In addition to highlighting the survival risks of pioneering really new product-markets, these empirical results help resolve an ongoing controversy in the order of market entry literature. The controversy is whether the first to market is the first to fail. In really new product-markets, the hazard rate modeling plots show that higher market pioneer survival risks tend to surface after six or more years in the market. This can arise when it takes a few years for a market pioneer to see if their initial investments, which were made in the face of great uncertainty, are on target. For incrementally new product-markets, the hazard rate modeling plots show that market pioneers have lower survival risks versus early followers. While these survival risk differences narrow over time, market pioneers have consistently lower survival risk. This can be explained by the pioneer’s initial monopoly, which increases survival rates in the first two to three years of commercialization. Also, pioneer first-mover advantages are often strongest in the early years of a market’s evolution and these advantages help protect the pioneer from outright failure. In conclusion, market pioneers are often the first to fail in really new product-markets. This is not true in incrementally new markets, where market pioneers have consistently lower survival risks versus early followers.
34 Research Limitations & Future Research
Four research limitations and the associated future research topics are discussed below. First, our sample is only composed of industrial goods markets. Different results could surface in different types of markets. For example, incremental innovations in high-technology markets may not generate meaningful pioneer survival benefits versus early followers. This is because a pioneer’s survival can be threatened by a higher degree of technological uncertainty. Second, our sample only contains 66 really new product-markets. While this sample size compares favorably with other survival studies, a larger sample size would yield more precise insights into market pioneer survival risks and how these survival risks change over time. Third, our product innovation newness measure is applied decades after the new markets started. Major investments in new products though arise prior to product launch, perhaps even at the prototype development stage. Thus, during the product development process, managers need a different type of measure to distinguish really new from incrementally new products. Fourth, optimal entry timing requires greater insights into fast followers versus slow followers. For example, how long should an early follower wait before attempting to enter a new market? While the Thomas Register data can address this research topic, it will require an expanded sample size plus an objective way to distinguish fast followers from slow followers.
Managerial Implications
Pioneering a new market is often considered to be a high risk investment with high potential returns. If so, we would expect to observe the highest returns to surface in the riskiest markets. At least from the standpoint of survival alone, this is not the case. This is because market pioneers have the lowest survival rates in really new product-markets, which carry the
35 highest risk. Given this pattern of results, why would anyone attempt to pioneer a really new product-market? A key reason is that if lower survival rates are compensated with higher profits, then pioneers might still be better off.5 Higher profits can arise because really new product-markets tend to be larger with higher profit margins (Mansfield and Wagner 1975). This indicates firstmover advantages can be especially strong for the few pioneering survivors. For instance, based on the Thomas Register data, in 1981 Hewlett-Packard, DatagraphiX, and IBM were the first laser printer entrants. While Hewlett-Packard is a dominant survivor, DatagraphiX exited in 1993. Thus, for really new products, while the risks of pioneering are high, so are the payoffs for the pioneers that do survive. Lower survival risks for early followers in really new product-markets suggest that early following can also be a profitable strategy. This is because not only do the early followers face a lower survival risk versus the market pioneers, they can also benefit from the market’s larger size and higher profit margins. These results are consistent with Olleros (1986) who contends that breakthrough innovations “are bound to be rather crude, costly, and unreliable when they first appear in the market (p. 11).” This provides an opportunity for early followers to leapfrog pioneers in really new product-markets. The risk and return profile in incremental innovation markets is more encouraging for market pioneers. Incremental innovation markets tend to have relatively low risk and higher pioneer survival rates. Incremental innovation markets are also likely to yield sustainable market share advantages for the pioneer. This is because research on surviving entrants points to
5
While profit data would yield valuable research insights, it is not provided in the Thomas Register. Profit data are available in PIMS, but the business unit and product category are both disguised. With disguised product categories, really new and incrementally new product markets cannot be identified. Thus, we do not know of any profit data that can be linked to really new versus incrementally new product-markets.
36 pioneer market share rewards. For examples, see Kalyanaram, Robinson, and Urban’s (1995) survey and VanderWerf and Mahon’s (1997) meta-analysis. With the majority of new markets being pioneered by incremental innovations, it is likely these market share reward studies are dominated by incremental innovation markets. If so, pioneers of incremental innovation markets should benefit from higher survival rates plus sustainable market share advantages.
37 TABLE 1 Product and Firm Exits, Mergers and Acquisitions, and Bankruptcies Between 1994-2002 Number of Product Exits
Number (Percent) of Product Exits That Were Accompanied by a Firm Exit
16 36
Market Pioneers Early Followers Total
Mergers or Acquisitions
Number of Bankruptcies
Number of Untraceable Cases
Survivors
Exits
4 (25%) 8 (22%)
0 1
1 0
0 2
3 5
44 82
12 (27%) 21 (26%)
1 0
1 6
2 0
8 15
178
45 (25%)
2
8
4
31
Really New Products Market Pioneers Early Followers Incrementally New Products
38 TABLE 2 Survey Questionnaire for Chemicals and Allied Products* ________________________________________________________________________________ Please evaluate whether each of the following products was a really new product or an incremental new product at the time of its initial commercial introduction. Note that: A really new product, or basic innovation, creates a new market, relies on new technology, and requires customer learning. An incremental new product, or modification is designed to satisfy a felt market need, and uses existing technology or refinement of it. Really Product Category Incremental New Carbon Monoxide Gas 1 2 3 4 5 6 7 Don’t Know DDT
1
2
3
4
5
6
7
Don’t Know
Dielectric Gas
1
2
3
4
5
6
7
Don’t Know
Chrome Dyes
1
2
3
4
5
6
7
Don’t Know
Indium Compounds
1
2
3
4
5
6
7
Don’t Know
Zinc Hardeners
1
2
3
4
5
6
7
Don’t Know
Lanthanides
1
2
3
4
5
6
7
Don’t Know
Saccharin
1
2
3
4
5
6
7
Don’t Know
Bromide Lead
1
2
3
4
5
6
7
Don’t Know
Lead Silico Fluoride
1
2
3
4
5
6
7
Don’t Know
Streptomycin
1
2
3
4
5
6
7
Don’t Know
Lead Titanate
1
2
3
4
5
6
7
Don’t Know
Lithium Acetate
1
2
3
4
5
6
7
Don’t Know
Lithium Niobate
1
2
3
4
5
6
7
Don’t Know
Magnesium Perchlorate
1
2
3
4
5
6
7
Don’t Know
Penicillin
1
2
3
4
5
6
7
Don’t Know
Manganous Sulfate
1
2
3
4
5
6
7
Don’t Know
*This is a part of the questionnaire for chemicals and allied products. The actual survey questionnaire includes 34 chemical and allied products.
39 TABLE 3 Inter-Experts Reliability: Cronbach’s Alpha Survey Group Industry (Two-Digit SIC Code)
Number Number of of Markets Raters
Cronbach’s Alpha
Textile Mills Products (22), Apparel and Fabrics (23), Lumber and Wood Products (24), Paper and Allied Products (26), Printing, Publishing, and Allied Industries (27) and Other Industrial Products
19
4
.95
Chemicals and Allied Products (28)
25
9
.87
Materials and Fabricated Metal Products (30, 32, 33, 34) – I
40
4
.84
Materials and Fabricated Metal Products (30, 32, 33, 34) – II
16
4
.96
Machinery and Computer Equipment (35) and Other Industrial Products – I
61
8
.90
Machinery and Computer Equipment (35) and Other Industrial Products – II
24
4
.93
Electronic and Electrical Equipment (36) – I
24
7
.88
Electronic and Electrical Equipment (36) – II
12
5
.86
Measuring, Analyzing, and Controlling Equipments (38) –I
18
8
.86
Measuring, Analyzing, and Controlling Equipments (38) –II
25
5
.76
26.40
5.64
.88
Average
40 TABLE 4 Sample by Two-Digit SIC Code (n=750) Really Incrementally New Products New Products
All
Industry (Two-Digit SIC Code) Textile Mills Products (22) Apparel and Fabrics (23) Lumber and Wood Products (24) Paper and Allied Products (26) Printing, Publishing, and Allied Industries (27) Chemicals and Allied Products (28) Rubber and Plastics Products (30) Stone, Clay, Glass, and Concrete Products (32) Primary Metal Industries (33) Fabricated Metal Products (34) Machinery and Computer Equipment (35) Electronic and Electrical Equipment (36) Measuring, Analyzing, and Controlling Equipments (38) Other Industries Total
0 0 0 0 0 13 3 0 2 2 13 12 17 4 66
3 3 5 4 2 12 19 8 5 17 62 24 26 8 198
3 3 5 4 2 25 22 8 7 19 75 36 43 12 264
Number of Firms Market Pioneers Early Followers Total
66 146 212
198 340 538
264 486 750
41 TABLE 5 Descriptive Statistics N
4-Year Survival
8-Year Survival
12-Year Survival
Market Pioneers (H1) Really New Products Incrementally New Products |t-Statistics|
66 198
76% 92% 3.72***
53% 79% 4.17***
23% 61% 5.62***
Early Followers (H2) Really New Products Incrementally New Products |t-Statistics|
146 340
74% 76% .45
50% 52% .47
38% 39% .32
Really New Products (H3) Market Pioneers Early Followers |t-Statistics|
66 146
76% 74% .27
53% 50% .41
23% 38% 2.17**
Incrementally New Products (H4) Market Pioneers 198 92% 79% 61% Early Followers 340 76% 52% 39% |t-Statistics| 5.53*** 6.64*** 5.05*** * p < .10, ** p
