The theoretical model in this study is tested on a sample of 203 SMEs. Since we were interested in internationally competitive SMEs, we first used a sampling criterion related to international sales, namely, annual exports of at least EUR 1 million. Second, since we were interested in the innovative behaviour of SMEs, we used the EU definition of SMEs (based on an employee headcount <250).1 Third, to reduce contextual biases and industry differences in innovation patenting, we decided to only include manufacturing firms. For the project, relatively labour-intensive on-site data collection was performed. The study was conducted in southern Sweden, a geographical area known for its entrepreneurial SMEs. To identify the sample, data were ordered from Statistics Sweden covering all firms in southern Sweden matching the sampling criteria. Two steps were then taken: first, the secondary data were evaluated in relation to the criteria; second, all firms were contacted by phone. At this stage, the researcher confirmed that the firms met the sampling criteria. After these two steps, the sample comprised 277 firms, 203 of which invited researchers to their facilities resulting in a response rate of 73%. The 74 firms that did not participate in the study were either unreachable after four attempts, declined to participate, said that they did not have time, or were uninterested. Post hoc, several non-response tests were conducted without revealing any significant differences or patterns when comparing responding and non-responding firms. The on-site data collection strategy was rather labour and cost intensive, but it provided the researcher with several advantages: it ensured respondent commitment, standardized the data collection procedure, and allowed the researcher to help respondents understand the questions. On each site visit, which lasted an average of 1.5h, a semi-structured interview and a standardized questionnaire were conducted. A template was developed for the visits to standardize the interview situation. To identify the respondents, an intra-firm snowball sampling process was followed. Although several meetings were conducted with two or three persons, the data are compiled and treated so that one person represents each firm. As this creates a risk of potential biases related to single respondents, the suggestions of Podsakoff, MacKenzie, Lee, and Podsakoff (2003) were carefully followed to avoid method biases common in behavioural research.

After the on-site data collection, we used objectively available register data on all 203 SMEs. From patent data registers in Sweden, we downloaded patent information for each firm. In addition, we downloaded data on each firm from official sources regarding firm age, ownership, growth, turnover, and profitability, so the dataset integrates survey data with data from secondary sources for each firm. Below is an account of how the individual constructs were measured and how the data for each variable were accessed.

As outlined above, measuring innovations is challenging, and the literature has suggested several approaches for doing so (see Taques et al., in press, for an overview). As we are mainly interested in the innovation output or performance of the firm, input-related measures such as R&D investment were not chosen. Instead, we followed the suggestions of Hagedoorn and Cloodt (2003) and Blomkvist, Kappen, and Zander (2017) and used patent data as a proxy for innovation. Since patent data enable us to create a reliable, replicable, and comparable dataset covering developments over time, we decided to download patent data for all sampled firms.

The independent variable of the model, time to innovation, is intended to capture the time elapsed from a firm's inception to its first innovation. Data were collected from Swedish patent registers on the first patent registered by each sampled firm. The year of firm foundation was then subtracted from the year of the firm's first patent; for example, a firm founded in 1993 with its first patent in 1998 has a time to innovation of five years.

The dependent variable of the model, pace of innovation, is intended to capture the innovativeness by the firm over time. For this purpose, data were collected from Swedish patent registers regarding the number of patents registered by each firm, which was divided by firm age. This measure indicates the average number of patents per year since firm inception; for example, a 10-year-old firm with eight patents has a value of 0.8.

The interacting variable of the model, networking, is intended to capture whether the firm actively seeks to tap its external network for capabilities to advance the innovation process. To capture this aspect, membership data were collected from the Swedish Chamber of Commerce. A dummy variable was created giving members a value of 1 and non-members a value of 0. The Swedish Chamber of Commerce has regional chapters in all counties of Sweden and actively strives to improve business conditions for its members, create networking opportunities among its members, and be a business hub in the regions where the chapters are located.

The regression model includes four control variables. First, firm size was controlled for by including the number of employees in the equation. Second, the international reach of the firm was controlled for by including a control for the number of markets to which the firm exports. Third, firm turnover was controlled for. Fourth, the share of the firms’ sales to foreign relative to domestic customers was controlled for.

Table 1 presents the descriptive data and the correlations between the variables of the model.

Ordinary least squares regression models were used to test the hypotheses, and Table 2 presents their results. Model 1 presents the results of the control variables, while model 2 adds time to innovation to the equation. Model 2 results indicate support for the first hypothesis, positing that the longer the time before first innovation, the slower the pace of innovation. This result is highly significant, with the adjusted R2 value indicating a predictive power of 27.8%. In model 3, the networking behaviour of the firm is added. The second hypothesis, positing that active networking by the firm moderates the negative relationship between time elapsed between firm inception and pace of innovation, is tested in model 3, which indicates significant support. The significant beta value changes from negative to positive, indicating that the networking effect can compensate for a late innovation start. The predictive power of model 3 is shown by the adjusted R2 value of 31.8%. These results are discussed in the next section.

This study has shown that the longer the time elapsed between a firm's inception and its first innovation, the slower the pace of innovation in the later development of the firm, emphasizing that timing matters for firm innovativeness. We argue that this finding indicates a liability of lateness in the innovative development of SMEs. By starting to develop innovations late, an SME will lag in building flexible and creative capabilities. This finding is supported by recent developments in international entrepreneurship research suggesting that firms that internationalize – which can be seen as an act of market innovation – can draw on the learning advantages of newness (Autio et al., 2000). It seems as though firms that are innovative at a younger age foster a culture and strategy that are open to exploring new ideas and to continuously developing new products in their growth efforts. Firms with a longer elapsed time between inception and first innovation, in contrast, display a slower pace of innovation in their continued development process. It seems as though they are hampered by the routines and structures developed during the period that elapsed before their first innovation. By looking at the underlying mechanisms of the relationships between the time when first innovations are registered and the continued innovations of the firm, we understand that the time taken to develop routines and capabilities matters. If technological capabilities are developed relatively late, the firm will have difficulties adapting these capabilities to changing market environments. Consequently, resource accumulation theory can help us understand the dynamics of innovation processes at the firm level. As shown, the negative effects of a late innovation start indicate that the development of innovation capabilities is hampered. The late start leads to a situation in which these firms fall behind the more active early innovators. In a rapidly changing business environment, this situation is arguably challenging for the late-starting firm. These firms risk losing market shares and competitiveness as the needs and wants of customers are likely to change before they begin innovating. In contrast, by starting the innovation process at a young age, SMEs can foster an innovative and open-minded organizational culture that constantly develops new ideas and products. A rapid pace of innovation will likely enable the firm to meet challenges arising in rapidly changing environments. A rapid pace of innovation also means that the firm will likely participate in changing the market environment, meaning that the firm is less reactive and is instead a proactive actor. By developing innovation capabilities at an early age, the firm is likely to drive market changes instead of being challenged by them.

Based on research into the development of innovation capabilities, one might assume that capability development is an internal process. However, by drawing on IMP research, we have shown that this is not necessarily the case. These findings are in line with recent suggestions that the relationship between technological capabilities and the internationalization performance of the firm is moderated by the social networks of managers (Zahra, 2019). Accordingly, we argue that there are problems with assuming that resource and capability development occurs within the firm. We have shown that firms that can access resources and capabilities in their external networks can compensate for the liability of a late innovation start. We further argue that accessing others’ resources and capabilities is a more sustainable strategy than seeking to develop them internally at high speed. Theoretically, we find clear explanations for this pattern in theories of time compression diseconomies (e.g., Dierickx & Cool, 1989; Hilmersson et al., 2017). Based on these findings, we question the importance of resource control and emphasize that critical resources may well reside in business networks. It can therefore be assumed that firms that cannot develop their own innovation capabilities, or that do so slowly, can compensate for their shortcomings by turning to their external networks. The network perspective is thus integrated with resource accumulation theory to advance our knowledge of the temporal aspect of innovation processes. Consequently, it was hypothesized that active networking by the firm moderates the negative relationship between time elapsed between firm inception and pace of innovation; the interaction test returned significant results, supporting the hypothesis. It can therefore be argued that resource accumulation theory falls short in explaining capability development processes in general and the innovation process in particular. The present findings reveal that resources and capabilities reside in business networks and that firms lacking certain resources and capabilities may well compensate for this lack by actively networking. This study therefore argues that networking is a means to accelerate the pace of firm innovation. These findings suggest that SMEs operating in rapidly changing environments where competitiveness is intense should seek to establish insidership positions (Johanson & Vahlne, 2009) in relevant business networks to accelerate their pace of innovation.

In light of Chen et al. (2012), who argued that the temporal aspect of innovation is understudied and deserves more explicit attention, this study has made an important contribution. First, the results contribute to the innovation literature by developing the concept of pace of innovation. Before this research, most literature on the temporal aspects of the innovation process focused on the innovation and development of certain products and their performance outcomes, using speed of innovation and time to market as concepts underlining the time it takes to develop certain products or services. Less was known about temporal innovation aspects from an SME management perspective. The present findings make a tentative contribution to the SME management literature by explicitly studying firm innovativeness and how to accelerate it over time. The measure developed captures the average number of innovations per year from firm inception instead of the time needed for certain innovations. Second, this study has made a tentative contribution to resource accumulation theory and, in particular, to innovation process theory by showing that resource accumulation is not necessarily an internal process. This research found that firms suffering from a late start in innovation capability development or that have underdeveloped innovation capabilities can alleviate such shortcomings by seeking resources and capabilities in their external networks. These findings validate research into the role of business networks in bolstering firm competitiveness (Dyer, 1996; Gulati, 1999; Zaheer & Bell, 2005). The present findings suggest that innovation research would benefit from further integrating the industrial marketing and purchasing perspective into its conceptualizations. This perspective builds on the idea that both buyers and sellers play active roles in business markets (Håkansson, 1982), as the resource constellation between firms is heterogeneous (Håkansson & Johanson, 1993). Firms can mobilize resources (Ritvala & Salmi, 2010) by actively networking with other business actors.