STRATEGIC ALLIANCES AND NEW PRODUCT DEVELOPMENT: AN EMPIRICAL STUDY OF THE U.S. SEMICONDUCTOR...

INTRODUCTION

A growing number of studies have begun to probe the impact of strategic alliance activities on technological learning and product development (e.g., Brown & Eisenhardt, 1997; Deeds & Hill, 1996; Kotabe & Swan, 1995; Lei, 1997; Mowery, Oxley & Silverman,

1996; Senker & Sharp, 1997; Steensma, 1996). This increased attention is long overdue given the importance of both strategic alliances (Barley, Feeman & Hybels, 1992; Contractor & Lorange, 1988; Gomes-Casseres, 1996; Hamel, Doz & Prahalad, 1989; Moore, 1996; Mowery et al., 1996; Nohria & Eccles, 1992; Ohmae, 1989; Teece, 1992); and new product development (Brown & Eisenhardt, 1995, 1997; Eisenhardt & Tabrizi, 1995; Kotabe & Swan, 1995; Lei, 1997).

So far, our knowledge of the linkage between strategic alliances and product development is still limited. Scholars disagree on whether strategic alliances enable firms to learn and to facilitate product development (e.g., Grant & Baden-Fuller, 1995; Kogut, 1988; Lei, 1997). Empirical studies also present a conflicting picture of the impact of strategic alliances on new product development (e.g., Brown & Eisenhardt, 1997; Deeds & Hill, 1996; Kotabe & Swan, 1995; Mowery et al., 1996).

To untangle the confusion in the extant literature, it is important to differentiate between different sources of learning in strategic alliances. A firm may learn from its alliance partners' organizational capabilities in different functional areas, such as new technologies, marketing skills, and manufacturing abilities (we call this content knowledge and learning). A firm may also gain knowledge from the management process of alliance activities (we call this process knowledge and learning). In the current study we used the difference between content and process learning as a departure point in an effort to extend our understanding of the impact of strategic alliances on new product development.(1)

Using US semiconductor industry startup firms in the 1980s, we empirically assess the impact of strategic alliances on new product development. This study finds that content learning in technological fields has a positive impact on product development while content learning in manufacturing and marketing areas shows no impact on product development. In addition, this study finds that process learning, measured by the comprehensiveness of alliance activities, appears to enable firms to gain valuable managerial process knowledge and in turn enhance their new product development.

The following section provides a literature review that forms the basis for the theoretical rationale and hypotheses developed in this paper. The method section details the study sample and its relevance and the research strategy. The results section presents empirical findings. The paper concludes with a discussion of the study's findings and future research directions.

In this research, strategic alliances are defined as interfirm collaboration with or without shared equity. Such activities include joint R&D, exchange of technology, second sourcing, manufacturing and marketing agreements or a combination of these. This definition, which excludes trade associations, interlocking directorship, mergers and acquisitions, and government-supported joint programs, such as SEMATECH (a semiconductor industry consortium), is consistent with that used in prior studies (e.g., Harrigan, 1986; Pisano and Teece, 1988; Shan, 1990).

LITERATURE REVIEW AND THEORY DEVELOPMENT

Previous studies on strategic alliances point out that turbulent external environments (e.g., Emery & Trist, 1965; McCann & Selsky, 1984), shrinking product life cycles, exploding R&D costs, and the increasing dispersion of skills and knowledge across firms and other organizations (Ohmae, 1989) have heightened the needs for strategic alliances (Gomes-Casseres, 1996; Hamel et al., 1989; Nohria & Eccles, 1992). At the same time, new product development has been rightfully placed as a central dimension of a firm's strategic processes and a vital part of its effectiveness (Brown & Eisenhardt, 1995; Eisenhardt & Tabrizi, 1995; Schilling & Hill, 1998). It is therefore natural for researchers to began to link strategic alliances and product development and expansion (Brown & Eisenhardt, 1997; Kotabe & Swan, 1995; Lei, 1997). However, from the emerging literature on this nexus, it is unclear whether or not strategic alliances will influence new product development, and if so in what fashion.

At a conceptual level, it remains unclear whether strategic alliances enable firms to learn from their partners. For example, Kogut (1988) points out that learning is not the only objective for forming strategic alliances; firms may engage in alliances to reduce transaction costs or maximize profits. In addition, Grant and Baden-Fuller (1995) argue that firms form alliances not to learn, but to use external resources so that they can concentrate on developing their existing capabilities. Theoretical predications in the strategic alliance literature, therefore, do not necessarily suggest that firms always use interfirm alliances to learn and to help develop new products. At the empirical level, several studies have looked at the relationship between a firm's alliance activities and their impact on product development. Mowery et al., (1996) find that strategic alliances make alliance partners' technology more similar, suggesting that firms do use alliances to learn from one another. Brown and Eisenhardt (1997) analyze six firms' product development processes, discovering that some firms used strategic alliances to probe the future; these effective probing practices in turn help those firms develop new products. At the same time the authors point out that one firm's alliance activities did not necessarily help its new product development; rather, its alliances were beneficial for extending the life of its existing products. Similarly, Kotabe and Swan (1995) fail to find support for the proposition that the number of alliances help product innovation. Instead, they argue that horizontal alliances (i.e., alliance partners that cooperate in the same functional areas) appear to help product innovation, while vertical alliances (i.e., alliance partners that cooperate in different functional areas) would not affect product innovation.

In sum, the existing literature is inconclusive regarding the relationship between strategic alliances and new product development. One reason, we believe, is that existing studies have not distinguished different sources of learning in strategic alliances. We believe a firm may learn through alliances' two types of knowledge: content and process knowledge. Content knowledge refers to the capabilities or skills in various functional areas. It may include technologies, manufacturing capabilities, marketing skills and other functional knowledge. Firms can access such content knowledge from alliances, which may in turn help product development.

A firm can also learn various managerial processes from alliance activities. In particular, a firm can learn two types of process knowledge. The first relates to managing interfirm cooperation. Many previous studies have pointed out that firms experienced in strategic alliances have learned to effectively negotiate cooperative contracts, manage inter-partner relations, and resolve cooperative difficulties (Doz, 1996; Hamel, 1991; Lyles, 1988; Ring & Van de Ven, 1994). While it doesn't appear to be directly related to product development, this type of process knowledge can have an indirect effect on new product development if we take the second type Of process learning into consideration.

Through alliances, firms may learn certain kinds of managerial process knowledge that can directly benefit their internal product development. First, a firm may learn to more effectively manage its cross-functional teams through its alliance activities and apply knowledge to its product development activities. Secondly, a firm may also learn to conduct environmental scanning more effectively through alliances, and use the knowledge to access and evaluate new product information. It is clear that this type of process knowledge can help a firm's product development. Both types of process knowledge, therefore, become central when considering a firm's new product development by means of strategic alliances.

Our theory development and hypothesis generation are based on the demarcation between content and process learning. We classify alliance content learning into three functional areas, namely, technological, manufacturing, and marketing; and we define alliance process learning as the extent to which strategic alliances are comprehensive in covering cross-functional activities. The next section explicates the effects of alliance content and process learning on product development.

Content Learning through Strategic Alliances

With respect to the content of interfirm learning, three functional areas are critical. They are technology, manufacturing capabilities, and marketing knowledge and skills. This study focuses on these three functional capabilities contributed by alliance partners as they constitute important elements in new product development (Daft, 1998; Porter, 1986). For example, Daft formulated the so-called horizontal linkage model of product development. This model identifies these three elements (technology, marketing and manufacturing) as key elements in product development, each charged with the task of monitoring its respective environmental segments as well (i.e., technological developments, customer needs, and manufacturing processes). With the guidance of this model, we explicate below the plausible impacts of learning from strategic alliances technologies, marketing, and manufacturing capabilities, on new product development.

As indicated earlier, scholars have identified multiple motives for strategic alliances. Some alliances may have content learning as their key motive while others may be more interested in resource and/or risk sharing, and specialization. These differing motives may provide some theoretical bases for understanding the impacts of learning different content knowledge on new product development. We next focus on the three different types of content learning.

Accessing technological capabilities. As mentioned before, the overall business environment has become increasingly turbulent (D'Aveni, 1994; Emery & Trist, 1965; McCann & Selsky, 1984). Coupled with this are a greatly reduced product life cycle, exploding R&D costs, the dispersion of technological skills and knowledge, and changing industry boundaries (Hamel et al., 1989; Ohmae, 1989). Under such circumstances, technological capabilities obtained through strategic alliances can help a firm develop new products. Three considerations support this assertion.

First of all, there is the benefit of shared R&D costs (Sakakibara, 1997). Secondly, there exists the possibility that a firm may be able to access knowledge and resources otherwise unavailable internally due to skill dispersion (Ohmae, 1989). In addition, if a firm is actively seeking technological capabilities through alliances, it may indicate a strategic intent that the firm is expanding its product-market domain. The basic literature in strategy indicates that accessing technology is fundamental to developing new products (Miles & Snow, 1978). Therefore we propose:

Hypothesis 1: The more actively a firm accesses technological capabilities through strategic alliances, the more likely the firm will develop new products.

Accessing manufacturing and marketing capabilities. While it is clear from the literature that accessing technological capabilities through strategic alliances will enhance product development, the respective impacts of accessing manufacturing and marketing capabilities are less conclusive. There are theoretical reasons for believing that access through strategic alliances manufacturing and marketing abilities may not necessarily be related to new product development. Specifically, manufacturing and marketing capabilities are different from technological capabilities. New manufacturing and marketing capabilities may help little if a firm does not have the requisite technological ability to develop new products. Furthermore, a firm's efforts to access manufacturing and marketing capabilities may be a clear indication that it wants to exploit its existing products rather than develop new ones. The empirical literature seems to support this notion as well. For example, in a study of strategic alliances and product development in the biochemical industry, Deeds and Hill (1996) exclude marketing alliances from their study. The authors argue that the purpose of marketing related alliances is not related to product development. In summary, this line of argument suggests that accessing manufacturing and marketing capabilities through strategic alliances would have no direct impact on product development. This argument is closely related to the resource sharing and specialization motives identified earlier.

Contrary to the above analysis, other studies indicate that new product development will benefit directly or indirectly from external manufacturing and marketing knowledge. Researchers in the product development field argue that interacting with outside knowledge sources helps product development. In an extensive review of the product development literature, Brown and Eisenhardt (1995) point out some evidence showing how communication with outsiders enhances a firm's product development. Von Hippel (1978) also indicates the importance of accessing market and customer knowledge in product development. Such market knowledge may help identify new customer demands and emerging market opportunities, which will in turn benefit a firm's product development. Similarly, external manufacturing may also help product development. A successful new product requires high quality and low cost manufacturing, and external manufacturing knowledge can help a firm achieve this manufacturing objective. In short, the above discussion suggests that accessing marketing and manufacturing capabilities through alliances can help new product development.

In addition, it is likely that firms may use alliances to access functional capabilities that they do not possess internally so that they can concentrate on developing their own distinctive competence (Grant & Baden-Fuller, 1995; Nakamura, Shaver & Yeung, 1996). This approach is not uncommon, especially among firms with limited resources. For example, in the semiconductor industry, many start-up firms choose to specialize in new product development and other technological activities (Dataquest, 1990a); they rely on strategic alliances and other means to access manufacturing and/or marketing resources. Overall, there appears to be some indication that accessing manufacturing capabilities and accessing marketing skills directly and indirectly help a firm's product development efforts. These arguments are related to both learning and resource sharing and the specialization motives identified above. If, indeed, firms are interested in accessing through strategic alliances content knowledge in marketing and manufacturing to advance their product development, accessing marketing and manufacturing knowledge will help new product development. We summarize these arguments as follows.

Hypothesis 2. The more active a firm accesses manufacturing capabilities through strategic alliances, the more likely that the firm will develop new products, and

Hypothesis 3. The more active a firm accesses marketing skills through strategic alliances, the more likely that the firm will develop new products.

Process Learning through Strategic Alliances

As outlined earlier, process knowledge gained through strategic alliances will be useful for a firm's alliance management and internal competence development. Through frequent strategic alliances, firms can not only learn to identify suitable partners for future cooperation but also manage interfirm collaboration more effectively. The interfirm collaboration, when managed effectively will avoid potential cultural clashes between cooperating firms, and more importantly enable them to benefit from complementary resources and capabilities. This itself becomes important for firms to utilize strategic alliances more successfully due to the complicated nature of alliance relationships (Parkhe, 1993). Furthermore, certain strategic alliances may require firms to use project teams across multiple functions. This cross-functional experience generated through alliances offers useful insights for a firm to utilize in its internal activities such as product development.

For firms to gain cross-functional knowledge through strategic alliances, the functional scope of its alliances, that is, the number of functional areas involved in an alliance, is crucial. The broader the functional areas a strategic alliance is involved in, the more likely that a firm will be able to derive valuable process knowledge useful for its product development. We outline our theoretical rationale below for this argument.

The theory of interdependence among organizational elements suggests that the more complicated the type of interdependence, the more involved the managerial process will be. From pooled to sequential and to reciprocal interdependence (Thompson, 1967), the managerial mechanics necessary to manage the interdependence are standards, policies, and team based coordination respectively. For firms that are engaged in more comprehensive alliances (alliances including more cross-functional activities), the interdependence among collaborators will most likely be more complicated, and more elaborating coordination mechanics would be necessary. This is because managing such interfirm cooperation presents much greater challenges than managing internal units (Parkhe, 1993). Because of this, firms that engage in comprehensive strategic alliances can gain insight on how to minimize potential cultural clashes among alliance partners and on how to utilize the respective skills and capabilities their partners possess. Such knowledge can benefit a firm's product development in two ways. First, as we hypothesized earlier, accessing different types of functional knowledge helps a firm's new product development. The broader the scope of alliances, the more effectively can a firm learn such content knowledge. For example, Hamel (1991), in a study of organizational learning in Japanese-Western alliances, indicates that multiple and cross-functional teams tend to increase the effectiveness of learning through strategic alliances, as such teams can learn new technologies and skills from multiple and different angles.

Secondly, alliances involving a greater number of functional areas will generate important insights in managing new product development processes (Brown & Eisenhardt, 1995; Dougherty, 1990, 1992). Successful product development involves effective R&D, manufacturing capabilities, and marketing skills. As Dougherty (1990) has demonstrated, various functional departments are tantamount to different thought worlds, each with its own stock of knowledge. Individuals in each area monitor and scan different parts of the environment (e.g., technological development, engineering processes, and customer trends) and understand different aspects of the product development process in different ways as well (Daft, 1998; Brown & Eisenhardt, 1995). Thus successful product development requires coordinating and integrating inputs from and cooperation of different functional areas. If a firm engages in comprehensive alliances, it will have more opportunities to learn how to create, organize, and manage cross-functional teams and attendant interactions among different areas. The managerial process developed from such alliances will enable a firm to manage more effectively its product development process.

In sum, a broad functional scope of alliances will help a firm's capability development in both interfirm collaboration and product development. It will enable a firm to learn more effectively of content knowledge that benefits product development; and more importantly, it will help a firm learn the skills of managing cross-functional fields, enhance its environmental scanning and sensing abilities, and apply such skills to its product development. Hence:

Hypothesis 4. The more functional areas a firm's alliances involve, the more likely that the firm will develop new products.

METHODS

Sample

The sample for the empirical assessments is comprised of US semiconductor start-up firms from 1978 to 1989. The database was compiled by Dataquest (cf., 1990a). Dataquest is an information company that specializes in the semiconductor industry. Its data are utilized extensively in both business reporting and academic research. Many business journals (from general outlets such as Fortune and Business Week to highly specialized ones such as Datamation and Information Week) rely upon Dataquest for important data in their reporting. Many academic articles also rely wholly or in part on data sets from Dataquest in their empirical analyses. These studies encompass strategic management topics (e.g., Boeker, 1997a, 1997b; Boeker & Goodstein, 1993; Burgelman, 1994; Eisenhardt & Schoonhoven, 1990), high technology management (e.g., Shi, 1998), quality (e.g., Sterman, Repenning and Kofman, 1997), marketing (e.g., Bayus, Jain & Rao, 1997; Brynjolfsson & Kemerer, 1996; Heide & Weiss, 1995), public administration (e.g., Kraemer, Gurbaxani & King, 1992).

The company compiled this database to track development in US semiconductor industry. According to Dataquest, the information in the database was gathered through multiple sources. Those sources range from self reports by companies, surveys, telephone interviews, to publicly available materials. The database contains detailed information on each firm, such as the founding date, product information, and interfirm cooperation activities.

Based on Dataquest's (1990a) definition, all the firms included in this database were less then ten years old, and having annual revenues not exceeding $100 million twice in their life time. All the firms design or ship finished products. The sample includes a total of 151 firms, and 95 new product introduction events during that time span. The sampling frame ended at 1989 for this current study, for that was the last year we began having access to the private data source. While a greater time span would have its advantages, we believe, as elaborated below, that there are strong reasons for choosing the 1978-1989 time frame for this current study.

First of all, the U.S. semiconductor industry experienced a great deal of change during this time span. The semiconductor markets saw a dramatic expansion. As Dataquest reported (1990a), the total industry sales at the beginning of the sampling time frame was $14.3 billion; by the end of 1989, the total industry sales reached $51.7 billion (1982 constant dollars). In addition, new technologies mushroomed and new products were being actively developed. For example, Application Specific Integrated Circuits (ASIC) products were developed in a short time span. In the memory chip area, fast SRAM (static random access memory), and EPROM (erasable programmable read only memory) products were also quickly developed during the time covered by this research. These developments represented an opportunity for startup firms to acquire new technology and expand into new product areas, and also a good opportunity to study start up firm's new product activities.

Secondly, the sample adopted in this study also allows us to overcome a problem common in prior studies. Several earlier studies focus on a firm's alliances within a period of time without considering its alliance history prior to the study period. This may lead to a left censoring bias as a firm's alliance activities at one time may have been affected by its prior strategic alliances. Focusing on start-up firms allows us to control the plausible longitudinal effect of alliance learning that might have occurred from a firm's inception.

In addition, Dataquest defines nine major product areas in the semiconductor industry (Dataquest, 1990b): ASIC, memory, analog, microcomponent, telecommunications ICs, discrete components, digital signal processing (DSP), gallium arsenide, and optoelectronics.(2) Within the confines of the semiconductor industry, these product areas differ in technical concepts, product design, and manufacturing capabilities. Firms do need different capabilities to produce different products. Hence, this industry provides an ideal setting for the current study. These three considerations support the use of this sample period for our study. A final consideration is relevant as well.

Clearly, the content of technological changes and resulting organizational activities reflected in this sample may be different from that of today. On the other hand, the managerial issues and lessons to be learned through this sample remain valuable to today's managers. This is because the current study examines the different types of learning taking place in strategic alliances and their subsequent impacts on product development; this issue has not received adequate attention in previous studies.

Measurements

Product expansion. The dependent variable for the current study is product expansion. Product expansion occurs when a firm adds a new product line or multiple product lines in a year. Though product expansion may not capture all product development activities, due to data limitation, we were able to examine only product development through expansion. This limitation will not cause major bias to our study, as it only makes our empirical study more conservative.(3) We created a binary variable to capture this construct. If a firm expands its product line in a given year, the dependent variable was coded as one; otherwise it was set to zero.

Below are the independent variables included in the empirical testing. To measure accessing different functional capabilities, we constructed variables related to the frequency and intensity by which a firm accessed technology, manufacturing, and marketing abilities respectively. The intensity and frequency measures capture the two dimensions of a firm's efforts to learn content knowledge. Including both dimensional measures allows us to test the arguments from different angles.

Total number of alliances accessing technology (Total Tech). This is the number of alliances in which a firm accessed technologies from its alliance partners. The technology may include receiving R&D and existing technologies from alliance partners. This variable measures the frequency of using alliances to access technology.

Average number of alliances accessing technology (Average Tech). This refers to the number of alliances through which a firm accessed technologies divided by the total number of alliances that firm had formed. This variable taps the intensity of a start-up firm to use alliances for technology acquisition purposes.

Total number of alliances accessing manufacturing (Total Mfg). This is the number of alliances in which a firm accessed manufacturing capabilities from its partners. It measures the frequency by which a firm accessed manufacturing capabilities from alliance activities.

Average number of alliances accessing manufacturing (Average mfg). This is the number of alliances in which a firm accessed manufacturing capabilities divided by the total number of alliances the firm had formed. This variable likewise measures the intensity of a start-up firm in using alliances to obtain manufacturing capabilities.

Total number of alliances accessing marketing (Total Mkt). This is the number of alliances in which a firm accessed marketing capabilities from its partners. Again, it taps the frequency with which a firm accessed marketing skills and abilities from its alliance activities.

Average number of alliances accessing marketing (Average Mkt). This is the number of alliances in which a firm accessed marketing abilities divided by the total number of alliances the firm had engaged. As before, this variable measures the intensity of a start-up firm to use alliances to access marketing capabilities.

Functional comprehensiveness of alliances (Comprehensive). To test the functional coverage or comprehensiveness of a firm's alliances, we measured the total number of functional areas (technological, manufacturing, and marketing) an alliance is involved in, regardless of which alliance partner is responsible for which functional activities. To capture the average functional comprehensiveness of a firm's alliances, we calculated the number of functional areas involved in each alliance the firm had formed, added that across all the alliances of the firm, and then divided this sum by the number of alliances the firm had engaged.

In addition to these hypothesis testing variables, we included several control variables in our analyses. They are as follows.

Total number of product lines (Product Lines). This is the total number of product lines a firm had. Theoretically, this construct may indicate a firm's internal capacities for developing additional products and should be controlled when studying the impact of alliances on new product development. Similar to the dependent variable, we utilized product lines as defined by Dataquest (i.e., ASIC, memory, analog, microcomponent, telecommunication device, discrete component, digital signal processing (DSP), gallium arsenide, and optoelectronics). A one year lag was used in the study.

Average amount of funding (Average Funding). This is the total amount of funding a firm received since its inception divided by the firm's age. Previous studies suggest that outside funding is crucial for start-up firms (Shan, 1990). This variable allows us to control the impact of a firm's financial resources on product development.

Age. This is the age of a start-up firm. This variable was used to control for the knowledge and experience a firm possesses with respect to new product development and competition, agility, and possible organizational motivation related to new product development.

Size. This is the total number of employees in a firm. As previous studies have pointed out, organizational size may affect a firm's adaptability and responsiveness (Hannah & Freeman, 1984); new product development certainly is a critical dimension of a firm's adaptability and responsiveness (Brown & Eisenhardt, 1995). The impact of size therefore should be controlled for in this study.

Manufacturing capability (Mfg Capability), A number of start-up firms in our sample did not possess manufacturing facilities. It is possible that a start-up firm may chose not to invest in manufacturing, and instead focus its resources on R&D and technological developments. Such a strategy may directly affect a firm's product development. It is therefore necessary to control for this aspect using a dummy variable.

Total Industry Sales Change (Sales Change). Finally, to capture the opportunities available in the semiconductor industry, we used total industry sales changes from year to year as a control variable.

Statistical Models

We created a binary dependent variable to tap a firm's product development and expansion. We used the Probit model that tests the impact of strategic alliances on product development. The Probit model has the form:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

[y.sub.i] = 1 if [z.sub.i] [is greater than] 0, and [y.sub.i] = 0 if [z.sub.i] [is less than or equal to] 0;

[[Epsilon].sub.i] ~ N(0,1).

LIMDEP, a statistic software package specialized in categorical and limited dependent variables, was used to estimate the Probit (Greene, 1995).(4) The Probit model is a nonlinear model using a set of independent variables to predict the probability that a certain event will occur (in our case, new product development). It is estimated using the maximum likelihood approach (Long, 1997). Each regression coefficient reflects a predicted probability change, that is, a change in the likelihood of a binary outcome (product development in our case) given a unit change in the associated independent variable. The overall level of significance for the model is assessed using the [chi square] statistic, and the significance of each regression coefficient is assessed through its associated z value.

RESEARCH FINDINGS

The empirical tests went through two stages. We first conducted a preliminary examination of the data set to ensure that it does not violate key assumptions of statistical models employed in the study. After this preliminary evaluation, we estimated various Probit models to test the hypotheses advanced in this paper. Table 1 details the descriptive statistics and correlation coefficients.

TABLE 1.
DESCRIPTIVE STATISTICS AND CORRELATION COEFFICIENTS

     Variables          Mean     s.d.      1       2       3       4

 1               Age     3.12     2.51    1.0
 2              Size   158.60   288.43    0.5     1.0
 3   Average Funding     2.84     6.05    0.2     0.4      1
 4    Mfg Capability     0.25     0.43    0.5     0.4     0.3     1.0
 5    Products Lines     1.42     0.93    0.2     0.3     0.1     0.2
 6      Sales Change     0.18     0.22   -0.8    -0.2    -0.3    -0.2
 7        Total Tech     0.66     1.39    0.4     0.36    0.3     0.37
 8      Average Tech     0.19     0.33    0.15    0.1     0.08    0.2
 9         Total Mfg     0.49     1.04    0.3     0.2     0.1     0.1
10       Average Mfg     0.14     0.27    0.09    0.03    0.00   -0.06
11         Total Mkt     0.23     0.59    0.3     0.38    0.3     0.26
12       Average Mkt     0.07     0.21    0.1     0.05    0.00    0.09
13     Comprehensive     0.84     0.82    0.3     0.2     0.2     0.2

     Variables          5       6      7      8      9      10     11

 1               Age
 2              Size
 3   Average Funding
 4    Mfg Capability
 5    Products Lines   1.0
 6      Sales Change   0.26    1.0
 7        Total Tech   0.34    0.01   1.0
 8      Average Tech   0.15    0.00   0.58   1.0
 9         Total Mfg   0.16    0.00   0.47   0.17   1.0
10       Average Mfg   0.08    0.00   0.14   0.11   0.7    1.0
11         Total Mkt   0.3    -0.0    0.53   0.16   0.39   0.2    1.0
12       Average Mkt   0.17   -0.02   0.08   0.1    0.12   0.3    0.6
13     Comprehensive   0.2     0.00   0.39   0.44   0.52   0.66   0.45

     Variables         12     13

 1               Age
 2              Size
 3   Average Funding
 4    Mfg Capability
 5    Products Lines
 6      Sales Change
 7        Total Tech
 8      Average Tech
 9         Total Mfg
10       Average Mfg
11         Total Mkt
12       Average Mkt   1.0
13     Comprehensive   0.5   1.00

Preliminary Examination

We conducted an assessment of multicollinearity using the conventional procedures of coefficient variance decomposition analysis with condition indices (SPSS Win, 7.5), and we found out that this data set did not present a multicollinearity problem, despite some relatively high correlation coefficients, specifically, no variable reached the conservative condition index threshold level of 15 prescribed in the method literature (Hair, Anderson, Tatham & Black, 1995). Furthermore, the data set is reasonably large by various guidelines including those for maximum likelihood (ML) estimations, thereby reducing the threat of collinearity (Long, 1997).

In addition, our data set is time series and cross-sectional or panel in nature (Sayrs, 1989). In such data sets, possible autocorrelation and heteroscedasticity may present another challenge (Greene, 1997). We therefore estimated the basic Probit model and the panel Probit model as implemented in LIMDEP (Greene, 1995), and we evaluated both sets of results carefully. As it turns out; the random effects for the panel models were significant for some models and insignificant for others. Furthermore, the basic model results were the same as that of the panel models. Because of this, we present only the basic model results in the paper.

Statistic Results

Table 2 summarizes the key research results based on the Probit models. We estimated five models. The first one (Model 1) includes only the control variables. Models 2 and 3 add three content learning variables conceived as frequency (Model 2) and intensity (Model 3) respectively. The last two models are extensions of Models 2 and 3 since we include the alliance functional comprehensiveness to probe the impact of process learning through alliances. The results related to two control variables are noteworthy. The average amount of funding received and the industry sales change positively affect product development. This suggests that a firm's internal resources and industry environment play key roles in product development.

TABLE 2.
PROBIT REGRESSION RESULTS(a)

Variables                  Model 1        Model 2        Model 3

              (Constant)   -2.0493(***)   -1.996(***)    -2.1938(***)
                           (0.15632)      (0.15884)      (0.1677)
                     Age    0.04719        0.0378         0.042478
                           (0.03114)      (0.03275)      (0.0325)
                     Sim    0.0001         0.0001         0.00018
                           (0.00019)      (0.00019)      (0.0002)
         Average Funding    0.03(***)      0.028(**)      0.031(***)
                           (0.0089)       (0.0093)       (0.00916)
          Mfg Capability   -0.04979       -0.12049       -0.13533
                           (0.16505)      (0.1734)       (0.17428)
           Product Lines    0.10266(*)     0.05528        0.05499
                           (0.0615)       (0.0663)       (0.0639)
            Sales Change    0.80668(**)    0.79876(***)   0.81445(***)
                           (0.28849)      (0.2904)       (0.294)

Hypothesis Testing
            Average Tech                                  0.68262(***)
                                                         (0.1755)
              Total Tech                   0.11019(**)
                                          (0.0498)
             Average Mfg                                  0.20719
                                                         (0.2314)
               Total Mfg                  -0.0124
                                          (0.0669)
             Average Mkt                                  0.32998
                                                         (0.2794)
               Total Mkt                  -0.030188
                                          (0.10799)
       Comprehensiveness

  Number of Observations   957            957            957
            [chi square]   35.72082       41.45386       53.96769
      Degrees of Freedom   6              9              9
      Model Significance   P < 0.001      P < 0.001      P < 0.0001

Variables                  Model 4       Model 5

              (Constant)   -2.23(***)    -2.265(***)
                           (0.18)        (0.1762)
                     Age    0.0344        0.032
                           (0.0339)      (0.0333)
                     Sim    0.00014       0.0002
                           (0.0002)      (0.0002)
         Average Funding    0.02604(**)   0.0267(***)
                           (0.0093)      (0.0095)
          Mfg Capability   -0.19968      -0.18187
                           (0.177)       (0.177)
           Product Lines    0.0296        0.0503
                           (0.0672)      (0.0643)
            Sales Change    0.814(***)    0.7995(***)
                           (0.2944)      (0.295)

Hypothesis Testing
            Average Tech                  0.49367(**)
                                         (0.203)
              Total Tech    0.11779(**)
                           (0.0496)
             Average Mfg                 -0.189
                                         (0.3137)
               Total Mfg   -0.10806
                           (0.0729)
             Average Mkt                 -0.042
                                         (0.343)
               Total Mkt   -0.10535
                           (0.109)
       Comprehensiveness    0.362(***)    0.28522(*)
                           (0.0965)      (0.1556)

  Number of Observations   957           957
            [chi square]   55.67595      57.28806
      Degrees of Freedom   10            10
      Model Significance   P < 0.0001    P < 0.0001

(a) (*) P < 0.1; (**): P < 0.05; (***): P < 0.01; standard errors are
in parentheses.

Of primary interest to this study are Models 4 and 5. The results show that content learning of technologies positively affects product development, and the impact is consistent, whether we measure that as a frequency or as an intensity. Thus, the result supports Hypothesis 1. On the other hand, the results fail to support Hypotheses 2 and 3 that accessing manufacturing or marketing skills help product development. Neither the frequency nor the intensity of learning manufacturing or marketing affects product development. Finally, functional comprehensiveness of alliances does affect product development in a positive fashion. Thus, Hypothesis 4 is supported. In general, our empirical results support hypotheses 1 and 4 that accessing technological capability (content learning in technology area) and engaging in comprehensive cross-functional alliances (process learning) help product development.

DISCUSSION AND CONCLUSIONS

This study began with the premises that strategic alliances will affect new product developments and expansion, and that learning through strategic alliances is related to both content and process. Based on our theoretical explication, we focused on the content learning in technology, manufacturing, and marketing respectively, and the process learning through functional comprehensiveness of strategic alliances. Using a longitudinal data set of U.S. semiconductor start-up firms from 1978 to 1989, the empirical results show that content learning in the technological field, and process learning from functional comprehensiveness in alliances positively affect product development; but content learning in manufacturing and marketing areas has no impact on product development.

Theoretical Implications

This study has important theoretical implications. While organizational learning through strategic alliances has received a great deal of attention in the literature, few studies have clarified different sources and effects of the learning. In this study, we argue that there are two sources from which firms can learn through alliances. The first source is alliance partners, in particular the capabilities or skills contributed by collaborative partners. The second learning source is alliance activities themselves, i.e., the management process of strategic alliances. These two learning sources create two types of learning, (a) learning the capabilities in certain functional areas, including technological, manufacturing and marketing skills, and capabilities (content learning); and (b) learning the managerial process (process learning). Within the process learning, some process knowledge relates to managing interfirm cooperation (e.g., how to negotiate and resolve interfirm conflicts), while other process knowledge can be applied to a firm's internal activities, such as product development.

Practical Implications

So far, existing studies have not paid adequate attention to process learning, especially to the fact that firms can apply alliance process knowledge to their internal activities. The distinction between different types of learning has important practical implications. It will enable companies to better understand where (what sources) they can seek organizational learning in alliances and what they can learn through those alliances. In addition, the two types of learning we discuss may be interrelated. As we indicate, cross-functional alliances may help firms learn technologies (content) and in turn benefit their product development. Since cross functional alliances relate to the structure of alliances that can be designed or organized by collaborative firms, firms should effectively design and organize their alliances so as to help them learn content knowledge more effectively from their alliance partners. Although cross-functional alliances may be more complicated to manage, our study suggests that companies that go through such alliances do gain a lot of organizational knowledge that can be transferred to their internal development. Therefore, complicated alliances do not always create difficulties; if managed carefully, they generate broader benefits to companies. In short, this research suggests that companies should not only focus on what they may learn through strategic alliances (content), they should also make efforts to benefit from the organizational process of interfirm collaboration.

Limitations and Future Research

There are several limitations in this study. First, this study focused on only the semiconductor start-up firms between 1978 and 1989. Because of the technological intensity of the industry and the unique characteristics of start-up firms, the generalizability of our findings remains an empirical question. Second, due to data limitation, this study was unable to look at several important product learning issues. For example, we were unable to directly examine the relationship between content and process learning. For process learning, we looked at only one type of process knowledge (cross-functional knowledge). It is not clear if there are other types of alliance process knowledge that may benefit firms' new product development directly.

With respect to future research directions, it is important to replicate this study in other empirical settings. The aforementioned weaknesses also require more attention in future studies. Another useful direction is the study of how learning in alliances with international partners might differ from learning with domestic partners. Psychological distances due to differences in cultural aspects (Hofstede, 1991) might be an important explanatory variable in such probes. Future studies may also look at the impact of content and process learning on other organizational behaviors such as innovation, and internationalization.

(1) The content and process learning discussed in the paper is different from "know-what" and "know how" concepts mentioned in other studies. For us, content learning is to learn functional knowledge (e.g., technological and manufacturing) from alliance partners, while process learning is to learn the process of engaging and managing cooperation and internal activities. So the content and process knowledge comes from different sources.

(2) According to Dataquest's definition (1990a: p. 74), ASICs are full or semi-custom integrated circuits. Memories are digitally integrated circuits used for data storage. Analog devices are integrated circuits in which data can assume a continuous, non-digital value. Microcomponents are semiconductor chips mainly used in personal computers, including graphics controllers, hard and floppy disk controllers, and microprocessors. Telecommunication ICs are semiconductor devices used in telecommunications equipment, such as satellites and facsimile machines. Discrete components are semiconductors containing single active elements, such as transistors and diodes. Gallium arsenide devices are integrated circuits in which the semiconductor material is gallium arsenide, not silicon. Finally, optoelectronics are light-producing and light sensitive devices, a category that includes mainly light-emitting devices (LEDs).

(3) Since our dependent variable did not capture all product development activities, it lost some empirical evidence that can be used to test our arguments. This makes it more difficult, rather than easier, to test the empirical relationship between alliance learning and product development. In this sense, the current dependent variable makes our empirical test more conservative.

(4) We used a dummy dependent variable and Probit models mainly because a majority of the firms expanded only one product line at a time. But in the analyses not reported here, we did code product expansion as a count dependent variable and used negative binomial models to test the same arguments (using the same set of independent variables). We find that results do not change with the way we code the dependent variable. To simplify the presentation, we did not report the results from the negative binomial model testing; they are available from the authors.

REFERENCES

Barley, S. R., Freeman, J., & Hybels, R. C. (1992). Strategic alliances in commercial biotechnology. In N. Nohria, & R. G. Eccles (Eds.), Networks and organizations (pp. 311-347). Boston, MA: Harvard Business School Press.

Bayus, B. L., Jain, S., & Rao, A. G. (1997). Too little, too early: Introduction timing and new product performance in the personal digital assistant industry. Journal of Marketing Research, 34, 50-63.

Boeker, W. (1997a). Strategic change: The influence of managerial characteristics and organizational growth. Academy of Management Journal, 40, 152-170.

Boeker, W. (1997b). Executive migration and strategic change: The effect of top manager movement on product-market entry. Administrative Science Quarterly, 42, 213-236.

Boeker, W., & Goodstein, J. (1993). Performance and successor choice: The moderating effects of governance and ownership. Academy of Management Journal, 36, 172-186.

Brown, S. L., & Eisenhardt, K. M. (1995). Product development: Past research, present findings, and future directions. Academy of Management Review, 20, 343-378.

Brown, S. L., & Eisenhardt, K. M. (1997). The art of continuous change: Linking complexity theory and time-paced evolution in relentlessly shifting organizations. Administrative Science Quarterly, 42, 1-34.

Brynjolfsson, E., & Kemerer, C. F. (1996). Network externalities in microcomputer software: An econometric analysis of the spreadsheet market. Management Science, 42, 1627-1647.

Burgelman, R. A. (1994). Fading memories: A process theory of strategic business exit in dynamic environments. Administrative Science Quarterly, 39, 24-56.

Contractor, F. J., & Lorange, P. (1988). Cooperative strategies in international business. Lexington, MA: Lexington Books.

D'Aveni, R. A. (1994). Hypercompetition: Managing the dynamics of strategic maneuvering. New York: The Free Press.

Daft, R. L. (1998). Organization theory and design (6th ed.). Cincinnati, OH: South Western College Publishing.

Dataquest. (1990a). A decade of semiconductor start-ups (3rd ed.). San Jose, CA: Author.

Dataquest. (1990b). Final worldwide semiconductor market share estimates. San Jose, CA: Author.

Deeds, D. L., & Hill, C. W. L. (1996). Strategic alliances and the rate of new product development: An empirical study of entrepreneurial biotechnology firms. Journal of Business Venturing, 11, 41-55.

Dougherty, D. (1992). A practice-centered model of organizational renewal through product innovation. Strategic Management Journal, 13(Summer), 77-92.

Dougherty, D. (1990). Understanding new markets for new products. Strategic Management Journal, 11(Summer), 59-78.

Doz, Y. L. (1996). The evolution of cooperation in strategic alliances: Initial conditions or learning processes? Strategic Management Journal, 17, 55-83.

Eisenhardt, K. M., & Schoonhoven, C. B. (1990). Organizational growth: Linking founding team, strategy, environment, and growth among U.S. semiconductor ventures, 1978-1988. Administrative Science Quarterly, 35, 504-529.

Eisenhardt, K. M., & Tabrizi, B. N. (1995). Accelerating adaptive processes: Product innovation in the global computer industry. Administrative Science Quarterly, 40, 84-110.

Emery, F. T., & Trist, E. L. (1965). The causal texture of organizational environments. Human Relations, 18, 21-31.

Gomes-Casseres, B. (1996). The alliance revolution: The new shape of business rivalry. Cambridge, MA: Harvard University Press.

Grant, R. M., & Baden-Fuller, C. (1995). A knowledge-based theory of inter-firm collaboration. Academy of Management Best Papers Proceedings: Academy of Management, Vancouver, Canada Academy of Management (pp. 17-21).

Greene, W. H. (1995). LIMDEP user's manual: Version 7.0. Bellport, NY: Econometric Software.

Greene, W. H. (1997). Econometric analysis (3rd ed.). Upper Saddle River, NJ: Prentice Hall.

Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1995). Multivariate data analysis with readings (4th ed.). Englewood Cliffs, NJ: Prentice Hall.

Hamel, G. (1991). Competition for competence and inter-partner learning within international strategic alliances. Strategic Management Journal, 12, 83-103.

Hamel, G., Doz, Y. L., & Prahalad, C. K. (1989). Collaborate with your competitors -- and win. Harvard Business Review, 67(1), 133-139.

Hannan, M. T., & Freeman, J. (1984). Structural inertia and organizational change. American Sociological Review, 49, 149-164.

Harrigan, K. R. (1986). Managing for joint venture success. Lexington, MA: Lexington.

Heide, J. B., & Weiss, A. M. (1995). Vendor consideration and switching behavior for buyers in high-technology markets. Journal of Marketing, 59, 30-43.

Hofstede, G. H. (1991). Cultures and organizations: Software of the mind. London: McGraw-Hill.

Kogut, B. (1988). Joint ventures: Theoretical and empirical perspectives. Strategic Management Journal, 9, 319-332.

Kotabe, M., & Swan, K. S. (1995). The role of strategic alliances in high-technology new product development. Strategic Management Journal, 16, 621-636.

Kraemer, K. L., Gurbaxani, V., & King, J. L. (1992). Economic development, government policy, and the diffusion of computing in Asia-Pacific countries. Public Administration Review, 52, 146-156.

Lei, D. T. (1997). Competence-building, technology fusion and competitive advantage: The key roles of organizational learning and strategic alliances. International Journal of Technology Management, 14, 208-237.

Long, J. S. (1997). Regression models for categorical and limited dependent variables. Thousand Oaks: Sage Publications.

Lyles, M. A. (1988). Learning among joint venture sophisticated firms. Management International Review, 28, 85-98.

McCann, J. E., & Selsky, J. (1984). Hyperturbulence and the emergence of type 5 environments. Academy of Management Review, 9, 460-470.

Miles, R. E., & Snow, C. C. (1978). Organizational strategy, structure, and process. New York: McGraw-Hill.

Moore, J. F. (1996). The death of competition: Leadership and strategy in the age of business ecosystem. New York: Harper Business.

Mowery, D. C., Oxley, J. E., & Silverman, B. S. (1996). Strategic alliances and interfirm knowledge transfer. Strategic Management Journal, 17, 77-91.

Nakamura, M., Shaver, J. M., & Yeung, B. (1996). An empirical investigation of joint venture dynamics: Evidence from U.S.-Japan joint ventures. International Journal of Industrial Organization, 14, 521-541.

Nohria, N., & Eccles, R. G. (Editor). (1992). Networks and organizations: Structure, form, and action. Boston, MA: Harvard Business School Press.

Norton, J. A., & Bass, F. M. (1992). Evolution of technological generations: The law of capture. Sloan Management Review, 33(2), 66-77.

Ohmae, K. (1989). The global logic of strategic alliances. Harvard Business Review, 67(2), 143-154.

Parkhe, A. (19931). "Messy" research, methodological predispositions, and theory development in international joint ventures. Academy of Management Review, 18, 227-268.

Pisano, G., Russo, M., & Teece, D. J. (1988). Joint ventures and collaborative arrangements in the telecommunications equipment industry. In D. C. Mowery (Ed.), International joint ventures in U.S. manufacturing. Cambridge, MA: Ballinger Publishing Company.

Porter, M. E. (Editor). (1986). Competition in global industries. Boston, MA: Harvard Business School Press.

Ring, P. S., & Van de Ven, A. H. (1994). Developmental processes of cooperative interorganizational relationships. Academy of Management Review, 19, 90-118.

Sakakibara, M. (1997). Heterogeneity of firm capabilities and cooperative research and development: An empirical examination of motives. Strategic Management Journal, 18, 143-164.

Sayrs, L. W. (1989). Pooled time series analysis. Newbury Park, CA: Sage Publications.

Schilling, M. A., & Hill, C. W. L. (1998). Managing the new product development process: Strategic imperatives. Academy of Management Executive, 12(3), 67-81.

Senker, J., & Sharp, M. (1997). Organizational learning in cooperative alliances: Some case studies in biotechnology. Technology Analysis & Strategic Management, 9, 35-51.

Shan, W. (1990). An empirical analysis of organizational strategies by entrepreneurial high-technology firms. Strategic Management Journal, 11, 129-139.

Shi, Y. (1998). The effects of tasks on semiconductor start-up alliance dissolution: A structuration perspective. Journal of High Technology Management Research, 9, 87-113.

Steensma, H. K. (1996). Acquiring technological competencies through inter-organization collaboration: An organizational learning perspective. Journal of Engineering & Technology Management, 12, 267-286.

Sterman, J. D., Repenning, N. P., & Kofman, F. (1997), Unanticipated side effects of successful quality programs: Exploring a paradox of organizational improvement. Management Science, 43, 503-521.

Teece, D. J. (1992). Competition, cooperation, and innovation: OrganizatiOnal arrangements for regimes of rapid technological progress. Journal of Economic Behavior & Organization, 18, 1-25.

Thompson, J. D. (1967). Organizations in action. New York: McGraw-Hill.

Von Hippel, E. (1978). A customer-active paradigm for industrial product idea generation. Research Policy, 7, 240-266.

Roger (Rongxin) Chen is Assistant Professor at University of San Francisco, California, USA; Mingfang Li is Associate Professor at California State University, Northridge, California, USA.