Financial resources for research and innovation in small and larger firms: Is it a case of the more you have, the more you do?

ABSTRACT Our study analyses how firms’ internal financial resources impact their engagement in scientific research, development, and five innovation activities. Furthermore, we investigate how firm-size moderates the impact of firms’ internal financial resources on scientific research, development, and innovation. Our approach provides novel insights regarding whether more money leads to more research and innovation, a topic that remains highly contested in the literature. Our analysis uses a novel unbalanced panel dataset of 1,446 firms in Ireland, over the period 2008–2016. Levels of internal financial resources are found to positively impact larger-sized firms’ (50+ employees) engagement in scientific research, process innovation and product innovation. However, such resources tend to hinder small-sized firms’ (less than 50 employees) engagement in service and organisational innovation. Our research refines innovation theory by reconciling contrasting views regarding the importance of financial resources for research and innovation, and offers novel insights for informing related public policy interventions.


Introduction
Behavioural and resource-based theories of innovation highlight the importance of firms' levels of internal financial resources for their engagement in research and innovation (R&I) (Cyert and March 1963;Jissink, Schweitzer, and Rohrbeck 2019;González-Bravo, López-Navarro, and Rey-Rocha 2021). The financial literature on innovation also proposes that financial constraints hinder firms' engagement in R&I (Himmelberg and Petersen 1994;Bond, Harhoff, and Van Reenen 2005;Hall et al. 2016). However, a growing number of studies, such as those of Berends et al. (2014), Bicen and Johnson (2014) and De Massis et al. (2018), demonstrate that some firms, especially small-sized firms, engage in R&I, despite having limited financial resources. Gibbert and Scranton (2009) and Keupp and Gassmann (2013), in particular, propose that financial resource constraints can encourage firm-level R&I. Therefore, as Weiss, Hoegl, and Gibbert (2017: p. 842) note, despite a large number of studies addressing this topic, the extent to which more money leads to more R&I remains 'unclear'. Mazursky (2008, p. 1389) have particularly emphasised that more research is needed to formulate a 'much-needed unifying theory of the role of financial resources in innovation management at large'. Keupp and Gassmann (2013) have also highlighted the specific need for a better understanding of innovation with limited financial resources. Developing a better understanding of this topic is more crucial now than ever (as we emerge from the Covid-19 pandemic), because it can inform the design and implementation of policies that support firms to innovate their way out of the current crisis (Roper 2020;Morgan et al. 2020).
Our paper addresses three critical intricacies regarding the impact of firms' levels of internal financial resources on their engagement in R&I, that have not heretofore, been fully explored in the literature. Our first contribution is to analyse how firms' internal financial resources impact their engagement in (i) scientific research and (ii) development. Arora, Belenzon, and Patacconi (2018) highlight that firms increasingly focus on exploiting existing knowledge, while their investments in new knowledge decline (i.e. explorative versus explorative research). They call for future research to address the 'underlying drivers' of such trends (Arora, Belenzon, and Patacconi 2018, p. 29). Studies up to now, have focused on financial resources as drivers of R&D investments (see, for example, Himmelberg and Petersen 1994; González-Bravo, López-Navarro, and Rey-Rocha 2021), but do not consider different types of research activities (Choi, Lee, and Bae 2019). Addressing this imbalance is important, as explorative research can enable firms to develop and sustain competitive advantage (Czarnitzki and Kraft 2012;Añón Higón 2016), and is also vital for addressing current social, environmental, and economic societal challenges (Borrás and Edler 2020). Our study makes a distinct contribution to the prevailing literature, by analysing how firms' levels of internal financial resources impact their engagement in scientific research (explorative) and development (exploitative). 1 From a policy perspective, our analysis is also important because it can usefully identify key entry points for innovation policy interventions to encourage more explorative research by firms (Czarnitzki, Hottenrott, and Thorwarth 2011).
Our second contribution is to unravel novel intricacies regarding firms' levels of internal financial resources impacting their engagement in five innovation activities: (i) process innovation; (ii) product innovation; (iii) service innovation; (iv) radical product and services innovation; and (v) organisational innovation. 2 Percival and Cozzarin (2008, p. 371) note that the prevailing literature on the drivers of innovation rarely considers the 'significance of the innovation under study'. Studies concerning the importance of financial resources for innovation, in particular, mainly focus on process and product innovation (Greve 2003;Keupp and Gassmann 2013;Pellegrino and Savona 2017;Giebel and Kraft 2019). This is a limitation that prevails in the literature, as firms may engage in different types of innovation activities depending on their levels of resources, including internal financial resources (Klingebiel and Rammer 2014).
Our paper extends the focus to innovation in services, which is an increasingly important innovation activity for service and manufacturing firms (Witell et al. 2017). As Mennens et al. (2018, p. 502) highlight, 'successful product and service innovation have different antecedents', such as their financial resource needs. We also consider organisational innovation, where 'R&D Sunk Cost' arguments of innovation may not apply (Tavassoli andKarlsson 2015, p. 1891). Furthermore, we distinguish between incremental and radical innovation activities, with the latter defined as products and services that are new to the market (Hewitt-Dundas, Gkypali, and Roper 2019). This is important, as the extent to which more financial resources lead to more radical forms of innovation remains a contested topic of research (see, for example, Weiss, Hoegl, and Gibbert 2017;Giebel and Kraft 2019). To the best of our knowledge, our study is the first to analyse the importance of internal financial resources for such a comprehensive set of innovation activities. Our analysis contributes to debates regarding the importance of financial resources for innovation, by going beyond whether more (less) money leads to more (less) innovation, and considering how firms' internal financial resources impact their probability of engaging in different forms of innovation. From a theoretical perspective, the insights of our study are vital for developing a more complete understanding of the importance of firms' financial resources for different forms of innovation.
Our third contribution is to unravel how firm size moderates the impact of firms' internal financial resources regarding their engagement in the seven R&I activities considered in this paper. We explore this by analysing the importance of internal financial resources for the R&I activities of small (10 to 49 employees) and medium and large-sized firms (i.e. hereafter referred to as larger-sized firms; 50+ employees), separately. 3 Focussing on small and larger-sized firms is a commonly adopted approach in the Irish context (See Hewitt-Dundas 2006;Vahter, Love, and Roper 2014;McGuirk, Lenihan, and Hart 2015, for examples), as most firms in Ireland are small-sized (98.9 percent), with larger-sized firms representing only 1.1 percent of firms (CSO 2020). The focus on small-sized firms is vital in the context of the prevailing literature concerning the impact of financial resources for R&I. This is because the available evidence points to a specific need to reconsider the conventional wisdom regarding the importance of financial resources for R&I in this firm-size group (Berends et al. 2014;Bicen and Johnson 2014;Colclough et al. 2019). Furthermore, the insights of our study provide new evidence that can be considered by policymakers, as they aim to encourage more R&I amongst firms of different sizes. The latter policy focus is seen in recent policy documents in varying country contexts (see, for example, Herr and Nettekoven 2018;OECD 2019;HM Government 2021).
In our study, to operationalise our analysis, we measure firms' internal financial resources as the ratio of Net Operating Surplus to turnover. This is similar to the measures of internal financial resources used by Czarnitzki, Hottenrott, and Thorwarth (2011), Hottenrott and Peters (2012) and González-Bravo, López-Navarro, and Rey-Rocha (2021). 4 Information on firms' R&I activities, and financial information, were obtained by merging a number of separate business surveys conducted by the Irish Central Statistics Office (CSO). Four waves of the Innovation in Irish Enterprises survey (IIE, formerly known as the Community Innovation Survey) were merged with four waves of the Business Expenditure on Research and Development (BERD) survey. The IIE dataset contains data on firms' engagement in innovation activities, while the BERD contains information on firms' engagement in scientific research and development. The Census of Industrial Production (CIP) and Annual Service Inquiry (ASI) surveys were also merged to obtain in-depth details of firms' internal resources, specifically their internal financial resources. This resulted in a novel unbalanced panel dataset of 2,531 observations from 1,446 firms in Ireland for the period 2008 to 2016.
Ireland represents an interesting locale for this study. Previous research has identified the availability of financial resources as an important determinant of the engagement in R&I activities by firms in Ireland (Roper and Arvanitis 2012), especially for small-sized firms (Hewitt-Dundas 2006). The period covered in this study also corresponds to the period following the 2008 Global Financial Crisis (GFC), when commercial lending significantly decreased, and firms' profits were the main source of funding for R&I by firms in Ireland (Central Bank of Ireland 2019). Recent data regarding the impact of the Covid-19 crisis show important similarities with the 2008 GFC (Central Statistics Office 2020). 5 As these business conditions are being experienced globally, the insights of our paper are particularly relevant in the current economic climate.
The remainder of the paper is organised as follows. Section 2 reviews the literature on the importance of firms' financial resources for their R&I activities and formulates hypotheses. Section 3 describes the data, the construction of the variables, and the empirical approach. Section 4 presents the empirical findings, which are then discussed in the context of the prevailing literature. Section 5 concludes by suggesting implications for the design and implementation of science and innovation policy interventions.

Literature review and hypotheses
Resource-based and behavioural theories of innovation regard the level of firms' internal financial resources as an important driver of their research and innovation (R&I) activities (Cyert and March 1963;González-Bravo, López-Navarro, and Rey-Rocha 2021). The financial literature on innovation proposes that firms may substitute internal financial resources with external finance (Hubbard 1998). However, market failures in financial markets may hinder firms' ability to access external finance for R&I, leading to firms typically relying on their internal financial resources for these activities (Himmelberg and Petersen 1994;Bond, Harhoff, and Van Reenen 2005;Hall et al. 2016). Some recent studies, such as González-Bravo, López-Navarro, andRey-Rocha (2021), De Massis et al. (2018) and Schäfer et al. (2017), have also demonstrated that firms typically prefer using internal financial resources for R&I, rather than incurring debt. Therefore, our focus on firms' levels of internal financial resources impacting their R&I activities is important. In this section, we review the literature on the importance of internal financial resources for R&I, and develop the hypotheses guiding the empirical analysis that follows.

Internal financial resources and scientific research and development
Firms typically face a trade-off between allocating limited R&D financial resources to explorative and/or exploitative research activities (Lee, Wu, and Pao 2014). However, Arora, Belenzon, and Patacconi (2018) demonstrate that firms increasingly favour exploitative over explorative research. This may lead firms to become increasingly dependent on external knowledge, and to curtail their ability to innovate in the future (Czarnitzki, Hottenrott, and Thorwarth 2011). Arora, Belenzon, and Patacconi (2018) specifically call for more research to unveil the factors driving these trends. We contribute to this topic by analysing how firms' levels of internal financial resources affect their engagement in scientific research and development.

Internal financial resources and scientific research
Scientific research is the 'R' of R&D (Czarnitzki, Hottenrott, and Thorwarth 2011), and focuses on generating logical explanations about physical, biological, and social phenomena by basic and applied research (Borrás and Edquist 2014). It enables firms to explore new ways of problem-solving for understanding and/or informing new technologies (Fleming and Sorenson 2004;Cassiman, Veugelers, and Arts 2018). This is important for building absorptive capacity, defined as a firm's ability to identify and benefit from external knowledge (Cohen and Levinthal 1990), and for avoiding technology lock-ins (Chadha 2011). However, scientific research may entail significant costs for hiring, training and retaining highly skilled employees (Arora, Belenzon, and Patacconi 2018). Most of these costs are sunk and highly volatile, as the returns on the investment may be lost if R&D employees leave the firm ). The knowledge generated by scientific research is also difficult to protect from other firms, and may not necessarily lead to commercial success (Dedrick and Kraemer 2015). The conjectural nature of scientific research, and its long-term focus, may lead firms to undervalue this activity (Aghion, David, and Foray 2009).
There are at least three mechanisms through which greater levels of financial resources can lead to firms engaging in scientific research. First, firms with high levels of internal financial resources can relax the expected returns of their research investments and engage in riskier research activities, because their internal financial resources can act as a safety net if these activities fail (Radas and Bozic 2012;Jissink, Schweitzer, and Rohrbeck 2019). Second, greater levels of internal financial resources can enable longerterm investments in research activities that enrich firms' knowledge breath, and potentially enhance their competence base (Lee, Wu, and Pao 2014). Finally, high levels of internal financial resources can ease firms' performance monitoring, and enable R&D employees to pursue non-core explorative research projects (Nohria and Gulati 1996;Jissink, Schweitzer, and Rohrbeck 2019). This suggests our first hypothesis: H1: Higher levels of firms' internal financial resources positively determine their probability of engaging in scientific research.

Internal financial resources and development
Development is the next stage, or the 'D', of R&D, and refers to the application of knowledge to, for instance, the development of innovations (Czarnitzki, Hottenrott, and Thorwarth 2011;OECD 2015). Development is an exploitative type of research, and firms engage in this activity to transform ideas and concepts into marketable technologies (Lee, Wu, and Pao 2014;Cassiman, Veugelers, and Arts 2018). Development can be resource intensive in terms of expenditure, but it is closer to the market and builds on existing knowledge (Czarnitzki, Hottenrott, and Thorwarth 2011). This means that firms may foresee potential returns to their investments, and reorganise their internal financial resources to finance development projects, especially those that are most valuable and near completion (Greve 2003;Berends et al. 2014). Firms can reduce the financial burden of development by engaging in collaboration (Grimpe and Sofka 2016;De Massis et al. 2018). They may also engage in new ways of value creation by recombining existing available resources, a process that Baker and Nelson (2005) term the 'bricolage' approach. This is similar to the notion of 'bounded creativity', as proposed by Hoegl, Gibbert, and Mazursky (2008). However, there are still significant financial demands for this activity, and internal financial resources are critical for financing development projects (Radas and Bozic 2012;Arora, Belenzon, and Patacconi 2018). This suggests: H2: Higher levels of firms' internal financial resources positively determine their probability of engaging in development.

Internal financial resources and process and incremental product and service innovation
We now consider the role of firms' internal financial resources as they pertain to process innovation, and incremental forms of innovation regarding products and services. Process innovation refers to improvements in production practices, such as the introduction of new technologies, methods of logistics, and maintenance systems (Tavassoli and Karlsson 2015). Incremental product innovation is defined as the introduction of goods that are significantly improved, or new to the firm, but that may already exist in the market (Berends et al. 2014). As process innovations may lead to new products, and new products may require new processes, their simultaneous consideration has significant benefits for firms (Percival and Cozzarin 2008;Hullova et al. 2019).
The literature supports the theory that internal financial resources are important for process and product innovation activities (De Falco and Renzi 2015;González-Bravo, López-Navarro, and Rey-Rocha 2021). Besides the costs associated with human capital, firms may need to obtain other tangible and intangible knowledge-creating resources when developing new processes and products, such as capital equipment, software, and licenced technology (D'Este, Amara, and Olmos-Peñuela 2016; Montresor and Vezzani 2016). The uncertainty of innovation, however, typically leads to firms financing innovation activities internally González-Bravo, López-Navarro, and Rey-Rocha 2021). Thus, greater levels of internal financial resources may facilitate firms to engage in more process and product innovation activities, and to dedicate more time and resources to these activities (Greve 2003;Jissink, Schweitzer, and Rohrbeck 2019). The prevailing empirical evidence supports the view that firms' levels of internal financial resources are important determinants for adopting new production technologies (Gomez and Vargas 2009), and developing new products (Nohria and Gulati 1996;Weiss, Hoegl, and Gibbert 2017). This suggests: H3: Higher levels of firms' internal financial resources positively determine their probability of engaging in process and product innovation.
Incremental service innovation refers to the introduction of services that are significantly improved, or new to the firm (Love, Roper, and Bryson 2011). While previous studies focused on service innovation in the context of service firms, manufacturing firms increasingly adopt service innovation as part of their competitive strategy (Witell et al. 2017). Firms engage in service innovation to improve productivity and increase sales (Hullova et al. 2019). Yet, service innovation departs from the technologically oriented nature of innovation in processes and products. Search strategies for service innovation are typically informal, and build on internal and external collaborations (Love, Roper, and Bryson 2011;Mennens et al. 2018). Firms may work closely with clients when innovating in services, thereby reducing uncertainty, and lowering the need for financial resources (Nijssen et al. 2006;Love, Roper, and Hewitt-Dundas 2010;Mennens et al. 2018). Mina, Bascavusoglu-Moreau, and Hughes (2014) propose that service innovation mainly arises from existing organisational and human resources, rather than from the availability of tangible assets. This suggests: H4: Firms' levels of internal financial resources do not determine their probability of engaging in service innovation.

Internal financial resources and radical innovation
In this section, we turn our attention to radical innovation, which we define as the development of products and services that are new to the market (Percival and Cozzarin 2008;Hewitt-Dundas, Gkypali, and Roper 2019). Earlier, in Section 2.2, we proposed that firms' levels of internal financial resources may have a differential causal effect between their engagement in product and service innovation. However, these differences may erode when innovation in products and services entails high levels of novelty (Mina, Bascavusoglu-Moreau, and Hughes 2014;Witell et al. 2017). Firms may face important knowledge discontinuities when bringing novel products and services to market, and their level of internal financial resources can impact their decision to engage in these activities (Percival and Cozzarin 2008).
There are two opposing arguments in the literature regarding the importance of financial resources for radical innovation. The first argument understands radical innovation as a function of its resource inputs (Weiss, Hoegl, and Gibbert 2017). Radical innovation is riskier than incremental innovation and more prone to fail, and internal financial resources are important for minimising these risks (Radas and Bozic 2012;Choi, Lee, and Bae 2019). Firms can experience knowledge discontinuities when innovating radically, and internal financial resources can enable firms to engage in more explorative and forward-looking search strategies to address them (Weiss, Hoegl, and Gibbert 2017;Jissink, Schweitzer, and Rohrbeck 2019;Choi, Lee, and Bae 2019). This suggests: H5: Higher levels of firms' internal financial resources positively determine their probability of engaging in radical innovation.
The second argument is in line with the notion of 'necessity is the mother of innovation' (Gibbert and Scranton 2009, p. 385). Limited financial resources may prompt firms to overcome knowledge discontinuities by deviating from existing path-dependencies, and find new cost-effective ways of doing things (Hoegl, Gibbert, and Mazursky 2008;Keupp and Gassmann 2013;Witell et al. 2017). They can find new ways of using their existing organisational resources, and/or exploit physical, social, or institutional inputs that other firms rejected or ignored (Baker and Nelson 2005). Furthermore, given that resources are limited, and that the potential rewards are high, managers may become more tolerant of riskier innovative projects that may improve their position in the market (Greve 2003). Thus, as proposed by Keupp and Gassmann (2013), firms' need to innovate (because of financial scarcity) can trigger new radical innovative efforts. This suggests: H6: Lower levels of firms' internal financial resources positively determine their probability of engaging in radical innovation.

Internal financial resources and organisational innovation
The final innovation activity considered is organisational innovation. This consists of new models of decision-making that support innovation by, for example, implementing new ways of managing relationships (within and outside the firm), and innovation portfolios (Birkinshaw, Hamel, and Mol 2008). According to Volberda, Van der Bosch, and Heij (2013), the drivers of organisational innovation remain under-researched. With specific relevance to our paper, the authors highlight that organisational innovation arises from changes within the firm, and may be unrelated to firms' levels of internal financial resources (Volberda, Van der Bosch, and Heij 2013). Furthermore, since this activity is neither R&D-based nor income-generating, resource-based arguments of innovation, such as the 'R&D sunk cost' argument, may not apply in this context (Tavassoli andKarlsson 2015, p. 1891). This suggests: H7 Firms' levels of internal financial resources do not determine their probability of engaging in organisational innovation.

Differences between small-sized and larger-sized firms
Finally, we focus on levels of internal financial resources having a heterogeneous effect on the research and innovation (R&I) activities of small (i.e. 10 to 49 employees) and largersized firms (50+ employees). Previous studies show that firm size is positively associated with the level of firms' internal financial resources (Freel 2000;Love and Roper 2015), and with their engagement in R&I (Himmelberg and Petersen 1994;Bond, Harhoff, and Van Reenen 2005;Hall et al. 2016). As a result, the literature tends typically to agree that small-sized firms' engagement in R&I is usually constrained by a lack of internal resources, especially financial resources (González-Bravo, López-Navarro, and Rey-Rocha 2021). However, higher levels of internal financial resources may not necessarily lead to more R&I by small-sized firms. This is because, as Berends et al. (2014: p. 618) highlight, theories of innovation mainly relate to large-sized firms but 'small firms are not miniature versions of large firms'.
Small-sized firms, especially those that are private and family-owned, may prefer growth and R&I strategies that preserve their long-term sustainability, as opposed to highly risky strategies (Schäfer, Stephan, and Mosquera 2017;De Massis et al. 2018;Croce, Grilli, and Murtinu 2019;Garrido-Prada et al. 2021). The competitive advantage of small-sized firms mainly resides in their behavioural resources, rather than in their capital and financial resources (Freel 2000). The flexibility of decision-making, for example, is a key behavioural advantage to refocus organisational objectives and routines when markets change (Berends et al. 2014). The high costs of R&I can hinder these behavioural advantages because the resources needed for these activities are largely sunk and not transferable to other areas (e.g. from research to marketing). Freel (2000), Bicen and Johnson (2014), and Berends et al. (2014) highlight that small-sized firms' R&I activities are typically informal, opportunistic, and more dependent on external knowledge and spillovers, rather than on internal R&D investments. On this basis, we examine the following and final hypothesis: H8: The positive impacts of firms' levels of internal financial resources on their R&I activities will be greater in larger-sized firms in comparison to small-sized firms.

Data and empirical approach
Progressing on from the conceptual development, this section describes the dataset, the construction of the variables, and the empirical strategy adopted in this paper.

Data
Our analysis uses a novel unbalanced panel dataset with information on firms in Ireland covering the period from 2008 to 2016. Our unbalanced panel merges several datasets from the Irish CSO. The Innovation in Irish Enterprises survey (IIE, formerly known as the Community Innovation Survey [CIS]) served as the base dataset. This is a biennial survey, with information on the internal characteristics and innovation activities of firms with at least 10 employees. Data from four survey waves (i.e. carried out in 2010, 2012, 2014 and 2016) were available to this study, covering a nine-year period from 2008 to 2016. To disaggregate firms' engagement in R&D into engagement in scientific research and engagement in development, the Business Expenditure in Research and Development (BERD) survey was merged with the IIE for the same period. This is a biennial survey containing detailed information on R&D expenditure by firms across all business sectors of the economy. The BERD survey indicates whether firms engaged in basic and applied (i.e. scientific), and experimental (i.e. development) research (See Section 3.2).
Both the IIE and BERD surveys are conducted every two years (2010, 2012, 2014, and 2016). In the IIE survey, the questions refer to firms' innovation activities during the previous three years (e.g. the 2012 IIE survey data provides information on innovation activities for the period 2010-2012). In the BERD survey, the questions refer to a twoyear period (e.g. the 2012 BERD survey data provide information on R&D activities for the period 2011-2012). 6 Section 3.2 provides details on the specific variables used. Moreover, the IIE survey is a representative sample of manufacturing firms, and firms in select services. 7 The BERD survey is 'designed to be a census of all enterprises that are believed to be engaged in research and development activities in all business sectors of the economy' (Central Statistics Office 2017, p. 4). 8 However, only around 45 percent of firms included in the BERD survey dataset typically engage in R&D across the different survey waves (e.g. 1,900 firms out of a sample of 4,200 firms performed R&D in the 2016 wave). Our available BERD data thus include information on R&D-active firms, and also on firms that do not engage in these activities. Some previous studies specifically focus on R&D-active firms when analysing the impact of financial constraints on firms' R&I investments (see, amongst many others, Bond, Harhoff, and Van Reenen 2005;Brown, Martinsson, and Petersen 2012). However, the focus of our study is on firms' engagement in R&I, and not just on firms' investments levels in research activities. As highlighted by González-Bravo, López-Navarro, and Rey-Rocha (2021, p. 8), it is 'necessary to consider both companies that do and do not engage in R&D' when studying the factors that drive firms to engage in R&I. Having data on frms that engage, and that do not engage, in R&D is thus appropriate in the context of our study.
Financial data to measure firms' levels of internal financial resources were obtained from the Census of Industrial Production (CIP) and the Annual Services Inquiry (ASI). These surveys are carried out every year by the CSO, and include data on firms' income and expenditure, thus allowing Net Operating Surplus, as the residual between income and expenditure, to be calculated (see Section 3.3). Both surveys are needed, as the CIP contains data on manufacturing firms, while the ASI focuses on firms in the service sector. Firms tend to base investment decisions on their internal financial resources generated in the previous period (Keupp and Gassmann 2013). Therefore, a-priori, one might expect a firm's decision to engage in research, development, or innovation in period t to be dependent on a firm's financial position in period t-1. For this reason, the financial data from the CIP and ASI data were merged with the IIE and BERD surveys to account for this lag. For example, the 2012 wave of the IIE and BERD datasets, covering the period 2010-2012 for the IIE and 2011-2012 for the BERD, were merged with the 2010 data in the CIP and ASI. 9 The final dataset, with the lag structure built-in, is an unbalanced panel with 2,531 observations from 1,446 firms. Around 45 percent of firms in the sample featured only once (674 firms), 50 percent featured at least twice (543 firms), 10 percent featured three times (147 firms), and 6 percent featured four times (83 firms). Around 63 percent of observations are from the 2014 and 2016 IIE and BERD survey waves (i.e. the two most recent waves available to this study). A total of 631 firms are small-sized, and 806 are larger-sized firms. Table 1 in Section 4.1 describes the sample in more detail.

Measuring research and innovation activity
Our study analyses the impact of firms' internal financial resources on their engagement in: (i) scientific research; (ii) development; (iii) process innovation; (iv) product innovation; (v) service innovation; (vi) radical goods and services innovation; and (vii) organisational innovation. Firms' engagement in each of these activities is measured in binary form. Scientific Research equals 1 if a firm records any expenditure on basic and/or applied research; otherwise, the value is zero. Development equals 1 if a firm invests in experimental research; otherwise, the value is zero. We consider basic and applied research activities as part of one indicator because both activities focus on generating new knowledge (Borrás and Edquist 2014). As Czarnitzki, Hottenrott, and Thorwarth (2011) explained, firms may engage in basic research when considering whether a concept and/or idea is viable in terms of fundamental scientific principles. Applied research, in turn, addresses potential challenges for the application of scientific concepts and/or ideas. These two research activities are fundamentally different from experimental research, which involves the process of adopting and scaling up the application of knowledge, including scientific knowledge, into products and services (Czarnitzki, Hottenrott, and Thorwarth 2011). The five innovation variables are also constructed in binary form, and take a value of 1 if a firm introduces the type of innovation considered, or otherwise zero. Panel A in Appendix A explains each variable.

Measuring internal financial resources
Internal Financial Resources is our key independent variable of interest, and measures the ratio of Net Operating Surplus (NOS) to total turnover. This is similar to the measures of internal financial resources used by numerous studies, including Czarnitzki (2006), Czarnitzki, Hottenrott, and Thorwarth (2011), Hottenrott and Peters (2012), and González-Bravo, López-Navarro, and Rey-Rocha (2021). NOS was obtained from the CIP and ASI datasets, by subtracting the following expenditure categories from total turnover: (i) total purchases of all goods and services other than capital items; (ii) changes in capital assets of the enterprise during the year; (iii) changes in intangible assets of the enterprise during the year; (iv) employment gross earnings; (v) other additional personnel costs (i.e. direct taxes and associated costs); and (vi) indirect taxes. We normalised by total turnover to account for heterogeneity arising from firm size (Schäfer, Stephan, and Mosquera 2017). Total turnover is directly declared by firms in both datasets. The resulting variable is continuous, with a possible maximum of 1 (all turnover is available as NOS) and a possible minimum of minus 1, as the data includes firms that exhibit losses.
While some literature focuses on financial slack (Jissink, Schweitzer, and Rohrbeck 2019) or cash-flows (for a review see Hall et al. 2016), such measures concern firms' internal and external financial resources, and require data on liabilities, which was not available to our study. As discussed in Section 2, we assume that firms tend to typically prefer financing their R&I activities internally, rather than incurring debt De Massis et al. 2018;Peia and Romelli 2020). Our available BERD survey data support this assumption. During the period covered in our study (i.e. 2008 to 2016), our BERD survey data show that internal financial resources were responsible for almost 90 percent of all firms' investments in these activities, with public support for R&D representing around 6 percent, and all other sources of funding (including external finance) representing the remaining 4 percent.
The financial literature on innovation proposes that firms may seek external finance when faced with potentially profitable investments, such as R&I activities (Hubbard 1998). However, the period covered in this study (i.e. 2008 to 2016) coincides with the period following the Global Financial Crisis (GFC) of 2008, where commercial lending significantly decreased in Ireland (Central Bank of Ireland 2019). Hoffmann and Sørensen (2015) explain that firms' access to credit was particularly limited in Ireland following the 2008 GFC, as domestic banks' inability to access international interbank credit led to stringent credit rationing. O'Toole, Gerlach-Kristen, and O'Connell (2013, p. 3) emphasise that credit constraints were especially severe for small and medium-sized enterprises (SMEs), that 'have traditionally been heavily reliant on bank-based lending'. Reports from the Irish CSO, based on the Access to Finance Survey (CSO 2011;, indicate that the percentage of enterprises seeking loans fell from 37 to 31 percent between 2007 and 2010, and less than 25 percent from 2012 to 2014. Loan success rates also ranged between 50 to 70 percent during most of this period, while the incidence of other sources of finance, such as equity finance, remained very low, at around 5 percent. 10 As presented in Table 1, our available BERD survey data indicates that only around 4 percent of the firms in our sample used some form of external finance for R&I from 2008 to 2010, with this percentage declining to around 2 percent during the 2012 to 2016 period. Profits were thus, the main source of finance for R&I by firms in Ireland during most of the period covered in our study. 11 According to Peia and Romelli (2020), tight credit supply, and uncertainty caused by the 2008 GFC, hampered firms' R&I activities in the European Union (EU), including Ireland, even if firms did not directly finance their R&I activities through debt. Therefore, although limited access to credit was particularly severe for firms in Ireland, firms in other EU countries faced similar circumstances Giebel and Kraft 2019;Peia and Romelli 2020). Furthermore, as the current Covid-19 pandemic is having similar effects in Ireland (Central Statistics Office 2020) and globally (Cowling, Brown, and Rocha 2020), our focus on firms' internal financial resources is highly relevant in the current context.

Additional independent variables
As is common in the literature, we controlled for firm size as the natural logarithm of the number of employees, and firms' levels of human capital as the percentage of employees that dedicate time to R&I (Shefer and Frenkel 2005;D'Este, Amara, and Olmos-Peñuela 2016). Consistent with studies using Irish data, such as Lenihan and Hart (2006) and Doran and Ryan (2014), we included a dummy variable to identify if firms are domestic or foreign-owned. A continuous variable measured the age of firms (in years), and accounts for possible heterogeneous effects between young and established firms. We also included a binary variable measuring if firms were exporters or not, as exporting is a widely recognised driver of firm-level research and innovation (Love and Roper 2015). Moreover, we included a dummy variable denoting whether firms were owned by an enterprise group or not, to control for potential access to resources, including financial resources, from the group (Jissink, Schweitzer, and Rohrbeck 2019). Firms can compensate for a lack of financial resources by accessing external knowledge through collaborations (Grimpe and Sofka 2016). Thus, we included four dummy variables capturing collaborations with: (i) clients; (ii) suppliers; (iii) other firms; and (iv) public knowledge providers.
Our data did not permit explicitly controlling for firms' access to finance for R&I, nor their leverage, which may influence their R&I activities (Hubbard 1998). However, we attempt to capture access to external finance for R&I in the following ways. The BERD survey required firms to indicate how they have financed their R&I activities, as follows: (i) internal/own financial resources; (ii) from other firms; (iii) Government grants; (iv) other public funding; (v) higher education institutes; (vi) private non-profit institutes; and (vii) other sources. The survey repeats the question to identify if the above funding sources are from Ireland, or from outside of Ireland. Given the specificity of these categories, firms likely declared using external finance in the 'other sources' category, as this is the only possible option in the survey to include such a source. 12 In light of the above, we construct four dummy variables, measuring whether: (a) firms received public financial support for R&I from the Irish Government; (b) firms received public financial support for R&I from the EU; (c) firms used external financial resources for R&I from Ireland; and (d) firms used external financial resources for R&I from outside of Ireland. The latter two variables (i.e. [c] and [d]) were constructed by grouping all external funding sources, as included in the BERD survey, apart from own funding, and government grants. While these variables do not explicitly account for the use of external finance for R&I, this is the best possible strategy permitted by our data. However, as discussed in Section 3.3, the vast majority of the firms in our sample primarily financed their R&I activities internally, and access to external finance was very limited for firms in Ireland during the period covered in our study (i.e. 2008 to 2016).
Finally, sectoral controls using one-digit NACE Rev. 2 classifications, and survey wave dummies (i.e. year), were included to control for sector and period effects as adopted in other studies such as Katila and Shane (2005). Panel B in Appendix A explains the construction of the additional independent variables.

Empirical approach
Our study focuses on unveiling causal relationships between firms' levels of internal financial resources and their engagement in research and innovation (R&I) activities. Hick (1980, p. 28) posited that establishing cause and effect relationships 'has to begin from some proposition, some relation between characteristics that has already been recognized'. Earlier in Section 2, we have unveiled the theoretical and logical mechanisms through which firms' levels of internal financial resources are expected to impact their engagement in the seven R&I activities considered, as emerging from the prevailing literature. Therefore, provided that our results support our hypotheses, we could interpret such results as causal 'on the basis of both logic and existing theory' (Lenihan, McGuirk, and Murphy 2019, p. 10). However, to ascertain causal relationships empirically, we need to consider issues of endogeneity that may influence the direction of causality (Reeb, Sakakibara, and Mahmood 2012), such as firms' financial profiles being determined by their previous innovative performance, and other unobserved factors (Hubbard 1998).
Previous similar studies to ours controlled for endogeneity by lagging suspected endogenous variables (see, for example, Czarnitzki 2006; Keupp and Gassmann 2013). Reed (2015) demonstrates that such an approach is not suitable in cross-sectional research designs, such as the one adopted in our study, as inter-temporal relationships between endogenous and outcome variables may remain. A valid solution to address endogeneity in the context of our study is to employ an instrumental variable approach (Czarnitzki and Kraft 2012). Here, suspected endogenous independent variables are instrumented by exogenous variables that: (i) correlate with the endogenous variable; and (ii) only influence the dependent variable through the endogenous variable (Reeb, Sakakibara, and Mahmood 2012;Montresor and Vezzani 2016). In the absence of suitable exogenous instruments, similar studies to ours used the lags of the suspected endogenous variables as instruments (see, for example, Tiwari et al. 2007;Borisova and Brown 2013;Lokshin and Mohnen 2013;Piekkola and Rahko 2019). Czarnitzki andKraft (2012, p. 1560) note that using the lags of suspected endogenous variables as instruments is 'common in the literature'. Reed (2015) demonstrates that this is an effective approach provided that: (a) the lagged instruments are sufficiently correlated with the suspected endogenous variable; and (b) do not themselves belong to the respective estimation equation.
Considering the above, we employed an instrumental variable approach using firms' lagged levels of internal financial resources, as an instrument for our main variable of interest Internal Financial Resources. Specifically, for each wave of the IIE and BERD surveys, from where we derived our dependent variables (i.e. t= 0), we used firms' average levels of internal financial resources for the years t-2 and t-3. For example, for the 2016 IIE and BERD survey wave (covering the period from 2014 to 2016), we used firms' average levels of internal financial resources for the periods 2011-2012 and 2012-2013 (which correspond to t-3 and t-2, respectively), as an instrument for our main variable of interest Internal Financial Resources. Using the two-year average was necessary to smooth changes in firms' internal financial resources, and obtain an instrument that is highly correlated with our suspected endogenous variable (i.e. the resulting correlation coefficient was 0.83). Moreover, our instrument does not belong to the contemporaneous equation, given that firms mainly base their R&I decisions on their immediate financial performance (i.e. t-1) rather on that of 2 to 3 years ago (Greve 2003;Jissink, Schweitzer, and Rohrbeck 2019). Our instrument thus meets the conditions as highlighted by Reed (2015).
We tested the validity of our instrument using regression-based tests, following Wooldridge (2010, p. 105). Such tests yielded F-statistics well above 10, which as Staiger and Stock (1997) have demonstrated, support the relevance of our instrument. Wald tests for endogeneity, however, only rejected the null hypothesis of exogeneity of the regressors in some of the R&I activities considered, and mainly in the context of larger-sized firms (see Table 4). This indicates that endogeneity was present in some, but not all, of the R&I activities considered. Therefore, as presented in Section 4.3, we repeated our analysis with two alternative estimators (i.e. random effect probit model and multivariate probit model), to ensure that our results were robust across different estimators and model specifications.
The impact of firms' levels of internal financial resources on their engagement in scientific research, development and innovation was estimated using the following equations: Equation (1) is an innovation production function (Geroski 1990) where IO it is a binary variable taking the value of 1 if firm i in period t engages in the research/innovation activity j under consideration, or zero otherwise. The model was estimated seven times, once each for: (i) scientific research; (ii) development; (iii) process innovation; (iv) product innovation; (v) service innovation; (vi) radical innovation in goods and services; and (vii) organisational innovation. The β term is the associated coefficient of interest for xIV it-1 , (i.e. the endogenous continuous variable Internal Financial Resources ranging from −1 to 1), which we instrumented by Equation (2). z itÀ 1 denotes the set of explanatory variables as discussed in Section 3.4, with ϕ denoting their associated coefficients.
Finally, ε it is the error term. In Equation (2), xIV itÀ 1 is our endogenous variable from Equation (1), which is linearly determined by our instrument x itÀ 2;3 and the same set of control variables as included in Equation (1). u it is the error term.
Equation (1) was estimated with an instrumental variable probit model with standard errors clustered at the industry level (i.e. one-digit NACE Rev. 2 classification), and Equation (2) with ordinary least squares. Both equations were estimated simultaneously. 13 The coefficient β in Equation (1) allows us to test hypotheses 1 through 7, where a positive and significant coefficient would provide support for hypotheses 1 to 3, and Hypothesis 5. Moreover, a negative and significant coefficient would provide support for Hypothesis 6, and non-significant coefficients would provide support to hypotheses 4 and 7. Section 3.6 details how we used the same model specifications to test Hypothesis 8 by firm sizes.

Financial resources in small-sized and larger-sized firms
Our study unravels the potential heterogeneous effects of financial resources between the engagement in research and innovation (R&I) by small (10 to 49 employees) and largersized (50+ employees) firms. Analysing medium (50 to 249 employees) together with large-sized (250+ employees) firms as one group is a commonly adopted approach in the Irish context (See Hewitt-Dundas 2006;Vahter, Love, and Roper 2014;McGuirk, Lenihan, and Hart 2015). This is because small-sized firms represent most firms in Ireland (98.9 percent), with larger-sized firms representing only 1.1 percent of firms (Central Statistics Office 2020). Likelihood ratio tests to assess parameter stability across these three disaggregated firm size categories confirmed that the estimates from the instrumental variable probit model were statistically similar to the sub-categories of medium and large-sized firms. However, they were significantly different for small-sized firms for most of the research and innovation activities considered (e.g. Chi2(22) = 83.06, P < 0.001 for scientific research).
A natural starting point to carry out the analysis as noted above, is to interact our continuous variable Internal Financial Resources with a binary variable denoting smallsized firms. This is appropriate in a model estimated via ordinary least squares (or another appropriate estimation technique). However, as outlined by Norton, Wang, and Ai (2004) and Karaca-Mandic, Norton, and Dowd (2012), the magnitude of interaction effects in non-linear models are conditional on the varying impact of the interacting variables (e.g. Internal Financial Resources and Small Firms), and all other explanatory variables. This makes interpreting interaction terms difficult in non-linear models, as the interaction effect cannot be summarised in one single coefficient (Norton, Wang, and Ai 2004). As probit models are a form of a non-linear model, this restriction applies to the current paper.
In light of the above, we followed Karaca-Mandic, Norton, and Dowd (2012) by repeating the analysis for each subgroup of firms. Equation (1) and Equation (2) were firstly estimated for a sub-sample of small-sized firms only. Separately, the same equations were estimated for a sub-sample of larger-sized firms. This provided comparable marginal effects of all the independent variables on firms' engagement in the various types of R&I activities for each size classification. Splitting the sample is standard in the literature on financial constraints, as firms' structural characteristics can influence their ability to both generate, and direct financial resources to R&I (see for example, Hottenrott, Hall, and Czarnitzki 2016). As noted above, Likelihood Ratio tests confirmed the stability of the parameters across these two firm-size groups. For completeness and robustness, we present a model including the interaction term in Section 4.3.

Empirical findings
This section reports the empirical findings. We focus on the average marginal effects for Internal Financial Resources, as this variable allows us to test our hypotheses. The coefficients for this variable measure the increase/decrease in the probability of firms engaging in each of the research and innovation (R&I) activities considered, given aoneunit change in their level of internal financial resources as aproportion of turnover. Before presenting our main results, we describe our sample and the variables used.

Descriptive statistics
As presented in Table 1, more than half of the firms in our sample engaged in at least one of the seven R&I activities, during each of the periods considered (i.e. 2010, 2012, 2014, and 2016). From Table 1, we also observe that a higher proportion of larger-sized firms engaged in the R&I activities considered, in comparison to small-sized ones. Correlation analysis indicates that firms typically engaged in more than one R&I activity during the same period. Around 47 percent of firms that engaged in scientific research, also engaged in development. Moreover, 62 percent of firms that engaged in product innovation, also engaged in radical innovation regarding products and services. Finally, 44 percent of firms that engaged in process innovation also engaged in organisational innovation. The correlation between other forms of R&I was generally low (i.e. below 0.3). These correlations are similar across the different survey wave years. Table 2 shows that, on average across the period 2008 to 2016, firms that engaged in most of the seven R&I activities had significantly higher levels of internal financial resources when compared to firms that did not engage in these activities (p < 0.05). This is apart from firms that engaged in service innovation, where the opposite is true (p < 0.01), and organisational innovations, where no differences are found. The table also shows that younger firms, and firms that engaged more in cooperation with clients, suppliers, and public knowledge providers, also engaged more in R&I (p < 0.05). The same occurred for exporting, and firms' levels of human capital (p < 0.05). Finally, firms that declared using external financial resources for R&I from outside of Ireland, generally engaged less in R&I (p < 0.05), but this was not the case for firms using external financial resources from Ireland, where no differences are found. As reported by Montresor and Vezzani (2022), firms may invest in other R&I related activities, such as branding and reputation building, and training. This may explain why firms that used external financial resources from abroad engaged less in the R&I activities considered in our study.

P-value
Diff.

P-value
Internal    Source: Authors' own elaboration using data from the Innovation in Irish Enterprises Survey (IIE), Business Expenditure on Research and Development (BERD) survey, Census of Industrial Production (CIP) and Annual Services Inquiry (ASI). Columns 'Diff' across the seven R&I activities indicate the existence of significant differences, and significance level, for each independent variable between firms that engage, and do not engage in the seven R&I activities considered (weighted by year). *** p < 0.01, ** p < 0.05, * p < 0.1.
Our descriptive analysis indicates that, on average, larger-sized firms had higher levels of internal financial resources than small-sized ones (around 2.5 percent [p < 0.01]). Larger-sized firms were also 2.8 years older (p < 0.01). Moreover, around 60 percent of larger-sized firms were owned by an enterprise group, compared to only around 20 percent amongst small-sized firms (p < 0.01). Small-sized firms, however, dedicated a higher share of their employee base to R&I activities, in comparison to their larger-sized firm counterparts (around 5 percent more [p < 0.01]). This is consistent with studies highlighting that small-sized firms tend to have a more informal approach to R&I activities, where not only R&D employees, but also other types of employees (such as manufacturing employees and firms' owners) contribute to such activities (see, for example, Freel 2000; Berends et al. 2014). Small-sized firms were also predominantly Irish-owned, while this was the case in only half of larger-sized firms (i.e. the difference between larger and small-sized firms is 27 percent on average, p < 0.01). Between 60 to 75 percent of smallsized firms exported, depending on the survey wave year considered (i.e. 2010, 2012, 2014, and 2016), while exports accounted for between 75 to 85 percent amongst largersized firms (i.e. the average difference, weighted by survey period, is 15 percent [p < 0.01]). Finally, around 4 percent of firms used external financial resources for R&I during the period from 2008 to 2010, declining to around 2 percent from 2010 to 2016. Such decline, however, was accompanied by an increase in the use of public financial support for R&I.

Main findings
Our main findings were obtained by estimating Equation (1) with an instrumental variable probit model, and are presented in Tables 3 and 4. 14 Wald-tests for exogeneity of the regressors are included in both tables. McFadden's pseudo R 2 s also indicate that the goodness of fit of our models ranged between 0.25 and 0.4, which according to McFadden (1977), indicates a good model fit.
Columns 1 and 2 of Table 3 show that firms' levels of internal financial resources affect their probability of engaging in scientific research and development differently. A onepercent increase in firms' internal financial resources (measured as net operating surplus divided by turnover) increases their probability of engaging in scientific research by 0.1 percent (Column 1). However, such resources do not determine firms' probability of engaging in development (Column 2). These results support Hypothesis 1, which posits that higher levels of firms' internal financial resources positively determine their engagement in scientific research, but do not support Hypothesis 2, which posits the same for development. As highlighted in the literature, our results may be attributed to the fact that scientific research is uncertain and requires long lead times (Fleming and Sorenson 2004;Cassiman, Veugelers, and Arts 2018;Mulligan et al. 2022). High levels of internal financial resources can allow firms to have more relaxed expectations regarding their explorative research investments (Lee, Wu, and Pao 2014;Jissink, Schweitzer, and Rohrbeck 2019). In contrast, firms may prioritise investments in development activities because they can lead to outcomes that are commercially viable in the short term (Czarnitzki, Hottenrott, and Thorwarth 2011;Arora, Belenzon, and Patacconi 2018). This is especially the case if firms focus on improving existing products 14 Estimations of Equation (2) can be made available upon request to the corresponding author. and services that are familiar to them (Czarnitzki, Hottenrott, and Thorwarth 2011;Radas and Bozic 2012). For these reasons, firms may engage in development activities regardless of their internal financial position. Organ. Innov.

Scientific Research
Dev.

Scientific Research
Dev.
Progressing on to our innovation indicators, Table 3 shows that firms' levels of internal financial resources have no impact on their engagement in process innovation and product innovation (Columns 3 and 4). Thus, our findings do not support Hypothesis 3, which posits that firms' internal financial resources positively determine their engagement in process and/or product innovation activities. Furthermore, our findings indicate that firms engage in service innovation when their levels of financial resources are low (Column 5). Here, an increase of one percent in firms' internal financial resources leads to a 0.04 percent decline in their probability of engaging in service innovation. Such findings do not support Hypothesis 4 that posits no relationship. Service innovation, therefore, may represent a way to innovation for firms with lower levels of internal financial resources (Mennens et al. 2018).
Similarly, our results in Column 6 in Table 2 do not support Hypotheses 5 and 6, regarding higher (and lower) levels of internal financial resources positively (and negatively) determining firms' engagement in radical innovation (regarding goods and services). These results may suggest that firms' engagement in radical innovation is mainly motivated by their ability to recombine existing resources, rather than the size of their financial resource base (Hoegl, Gibbert, and Mazursky 2008;Gibbert and Scranton 2009;Bicen and Johnson 2014;Colclough et al. 2019). Colclough et al. (2019), for example, highlight that firms' financial resources may not be an essential determinant of their engagement in radical innovation. In their view, this type of innovation is primarily driven by firms' innovation orientation, which is determined by the growth ambitions of firms' managers and owners. Our findings support these arguments. In contrast, however, our findings support Hypothesis 7, as levels of financial resources have no impact on firms' engagement in organisational innovation (Column 7). This innovation activity centres on reorganising routines for R&I, and it may be the case that firms rely on their existing human capital resources and/or may collaborate with external partners when carrying out these changes (Volberda, Van der Bosch, and Heij 2013;Tavassoli and Karlsson 2015).
Our analysis unravels significant differences when it is carried out for small, and larger-sized firms separately. Table 4 shows that the positive relationship between firms' levels of internal financial resources and their probability of engaging in scientific research only applies in the context of larger-sized firms (Column 1 in Panel B). Regarding innovation activities, we do not find that financial resources encourage innovation in the case of small-sized firms. Panel A also shows that small firms' levels of internal financial resources negatively determine their probability of engaging in service and organisational innovation (Columns 5 and 7). In contrast, a one-percent increase in firms' internal financial resources increases the probability of larger-sized firms engaging in process and product innovation by 0.2 percent (Columns 3 and 4 in Panel B). These results support to some extent Hypothesis 3, but only for this firm-size group, and are consistent with resource-based and behavioural theories of innovation (Nohria and Gulati 1996;González-Bravo, López-Navarro, and Rey-Rocha 2021). Therefore, taken together, our results support Hypothesis 8, which proposes that the impact of firms' levels of financial resources on their engagement in R&I will be greater in the context of larger-sized firms, when compared to small-sized firms.
As firm size typically correlates with the level of firms' financial and human capital resources (Freel 2000), larger-sized firms may be more able to engage in scientific research than small-sized firms (Arora, Belenzon, and Patacconi 2018). Size naturally represents a limit to the number of research projects that firms can undertake, and smallsized firms may concentrate on fewer, but higher-quality projects (Greve 2003;Berends et al. 2014). Thus, they may refrain from investing additional financial resources in research activities when their R&D resources are fully occupied (Berends et al. 2014). Moreover, small-sized firms may prefer research projects that focus on applying their existing expertise, rather than projects that require expanding their expertise, especially when operating in niche markets where they hold knowledge advantages (De Massis et al. 2018).
Regarding innovation activities, González-Bravo, López-Navarro, and Rey-Rocha (2021) suggest that the availability of resources in absolute terms may not be as important for innovation as firms' ability to generate them efficiently. Thus, small market presence and uncertain international markets, may deter small-sized firms from engaging in more innovation, even if they have sufficient financial resources (Berends et al. 2014;De Massis et al. 2018). Added to this, managers in small-sized firms may engage in some forms of innovation, such as new services, as a means of generating new revenue streams when their levels of financial resources are low (Greve 2003;Mennens et al. 2018). Equally, they may focus on implementing organisational changes as a means of becoming more efficient at innovating (Volberda, Van der Bosch, and Heij 2013). In this sense, financial scarcity may trigger new innovative efforts by small-sized firms (Keupp and Gassmann 2013). Our results, therefore, concur with authors such as Bicen and Johnson (2014) and Berends et al. (2014), when proposing that the conventional wisdom regarding the importance of financial resources for innovation may not apply to small-sized firms.
The control variables show that firm size, the number of R&D employees, and exporting all drive small and larger-sized firms to engage in all R&I activities considered. Furthermore, both firm-size groups benefit from cooperation in a similar manner. Cooperation with public knowledge providers (i.e. research centres and universities) drives both firm-size groups to engage in scientific research and development. Cooperating with clients also drives firms to engage in development, and most cooperation variables are positive and significant for all innovation activities. Larger-sized firms that are domestically-owned, are more likely to engage in scientific research, and product innovation, than their foreign-owned firm counterparts. Our findings concur with those of Roper and Arvanitis (2012) who found that Irish-owned firms may be more likely to source knowledge internally, by investing in R&D, in comparison to foreign-owned firms. Our findings, however, differ from those of McGuirk, Lenihan, and Hart (2015), who found Irish-owned firms to be less likely to innovate when compared to foreignowned firms. We attribute such difference to our sample, which comprises firms that are believed to perform R&D activities, as identified by the Irish CSO, while McGuirk, Lenihan, and Hart (2015) analysed data comprising of a sample of all firms in Ireland from a workplace survey (i.e. the Irish National Centre for Partnership and Performance [NCPP] 2009 Workplace Survey).
Finally, public financial support for R&I increases the probability of engaging in all activities considered for both firm-size groups. Moreover, the impact of external sources of finance depends on whether such sources are from Ireland or from abroad, and on the size of the firms. That is, obtaining financial resources from domestic sources increases small-sized firms' probability of engaging in scientific research and development activities. External sources of finance from domestic sources increases larger-sized firms probability of engaging in radical innovation. However, obtaining financial resources from abroad negatively impact firms' engagement in all R&I activities considered, by both firm-size group. As noted in Section 4.1, firms may use external financial resources for specific R&I-related activities, such as branding and reputation building, but may mainly finance their R&I activities internally.

Robustness checks
To ensure that our results are robust to reasonable alternative specifications of the model, we replicated the analysis using different specifications of Equation (1), and by estimating the full equation with different estimators. In all cases, the direction and significance of the impacts identified by our main analysis were supported by our robustness checks.
To ensure that our results are not affected by multicollinearity, we re-estimated Equation (1), but excluded some of the variables from our original model in the analysis. That is, Equation (1) was firstly re-estimated (for the whole sample), but only included the variables measuring firms' characteristics. We repeated this approach by estimating Equation (1), but only included the variables measuring firms' characteristics, and external sources of funding (i.e. excluding the cooperation variables). As presented in Appendix B, excluding variables from our model did not affect our results. Secondly, we re-estimated Equation (1) with a random effect estimator to account for the panel dimension of the data (i.e. within-firm heterogeneity). We expected this specification to control for inter-temporalities between firms' financial resources and their R&I activities. Appendix C shows that the results obtained with a random effect probit model for small and larger-sized firms support the direction and significance of the results obtained by the instrumental variable probit estimator. Some small differences in the coefficients are observed, which is as expected, given the different functional forms.
Thirdly, as firms may engage in a portfolio of R&I activities at a given time (Klingebiel and Rammer 2014), we re-estimated Equation (1), but now considered the correlation between different R&I activities with a multivariate probit model (Galia and Legros 2004). Following Roper, Du, and Love (2008), the efficiency gains derived from a multivariate probit model are limited when the vector of explanatory variables are strongly correlated across the different outcomes. This was expected here, given that these variables capture firms' internal and external determinants of the different types of science and innovation activities considered in our study. Appendix D thus confirms that the results of the multivariate probit model are very similar to our main findings as presented in Table 4.
Finally, we re-estimated Equation (1) for the whole sample, but now included an interaction term between the continuous variable Internal Financial Resources and a dummy variable that equals 1 if the firm is small-sized (10 to 49 employees), or zero otherwise. In this specification, the variable Internal Financial Resources measures the effect of internal financial resources on larger-sized firms' probability of engaging in the research/innovation activity under consideration. The interaction variable does the same for small-sized firms. Following Norton, Wang, and Ai (2004), we present average marginal effects for the interaction term computed as the full derivative, instead of the partial derivative or partial effect. 15 Appendix E shows that the results of this specification largely support our main results, with the only difference being that the coefficient for Internal Financial Resources is now positive and significant for development in the context of larger-sized firms. This may be due to the increased precision of the standard errors, given the larger sample size. Our robustness checks indicate that our main results obtained with the instrumental variable probit model are robust across different model specifications and econometric approaches.

Conclusions and implications for policy
Our paper compares the impact of levels of internal financial resources between small (10 to 49 employees) and larger-sized (50+ employees) firms' engagement in scientific research, development, and five types of innovation activities. To the best of our knowledge, this is the first study to provide such a detailed analysis regarding the impact of firms' internal financial resources on their research and innovation (R&I) activities. Our study, therefore, extends previous studies that mainly focused on R&D, process innovation and product innovation (De Falco and Renzi 2015; González-Bravo, López-Navarro, and Rey-Rocha 2021). Our findings shed new light on a topic that remains highly contested in the prevailing literature (Weiss, Hoegl, and Gibbert 2017), by providing a more nuanced understanding of the 'significance of the innovation under study' (Percival and Cozzarin 2008, p. 371). These insights are crucial for the formulation of a 'much-needed unifying theory of the role of financial resources in innovation management at large' Mazursky 2008, p. 1389).
From a theoretical perspective, our research provides further insights into three key issues that remain largely unresolved in the prevailing literature. First, it highlights the importance of considering firms' levels of internal financial resources for understanding their increasing focus on development, while their investments in scientific research decline (Arora, Belenzon, and Patacconi 2018). As scientific research has long lead times and unexpected outcomes, firms may undervalue this research activity (Aghion, David, and Foray 2009). Thus, they may only engage in scientific research when they have high levels of financial resources (Jissink, Schweitzer, and Rohrbeck 2019). In contrast, development is closer to the market, and firms may engage in this activity regardless of their financial position, as they may foresee concrete returns to their investments (Czarnitzki, Hottenrott, and Thorwarth 2011).
Second, our paper demonstrates that it is important to extend the focus beyond process and product innovation when analysing the role of financial resources for firmlevel innovation. In line with behavioural and resource-based theories, we find that financial resources drive firms to engage more in process and product innovation, but only in the context of larger-sized firms. However, firms may not need high levels of financial resources for other types of innovation, such as service, organisational and radical innovation (regarding goods and services). Here, firms may favour 'bricolage' (Baker and Nelson 2005) or 'bounded creativity' (Hoegl, Gibbert, and Mazursky 2008) approaches, where the recombination of existing resources in new ways can mitigate the need for financial resources. Finally, our research unravels important differences regarding the influence of financial resources on the R&I activities of small and largersized firms. In particular, our results suggest that small-sized firms may view R&I as a counter-measure to when their performance in the market falls, as opposed to an opportunity to grow, which is consistent with arguments of financial scarcity leading to more R&I (Hoegl, Gibbert, and Mazursky 2008;Keupp and Gassmann 2013;Witell et al. 2017).
Our analysis uses data for firms in Ireland for the period following the 2008 Global Financial Crisis, which has similarities with the current Covid-19 crisis (CSO 2020). Therefore, the insights arising from this paper are highly topical as they can elucidate the potential impact of the Covid-19 pandemic on firms' research and innovative activities through their effects on firms' financial resources. The insights provided here can serve as a platform for discussions and analyses regarding the design and implementation of more targeted science and innovation policy interventions to help firms to innovate their way out of the current Covid-19 crisis (Roper 2020;Morgan et al. 2020). In particular, the prevailing literature widely regards public financial support as a critical innovation policy intervention to promote R&I by small-sized firms . Our results suggest that such support may be complemented with other types of government intervention (i.e. beyond financial instruments) that specifically target, for example, the development of innovative human capital, and other means to inspire and enable small-sized firms to engage more in these activities (McGuirk, Lenihan, and Hart 2015;Lenihan, McGuirk, and Murphy 2019). Policies to accelerate new models of collaboration amongst small-sized firms, and with public knowledge providers, can further support these firms to innovate (Hewitt-Dundas and Roper 2018). Though beyond the scope of the current paper, these are certainly areas worthy of further exploration and investigation.
Regarding larger-sized firms, the positive relationship between their level of internal financial resources and their engagement in scientific research may signal policy entry points for governments who want to encourage firms to engage more in this activity. Public financial support may further stimulate investment in scientific research activities . This support, however, should target scientific research projects with high social rates of return to avoid substituting private investments (Mazzucato 2016), and thus avoiding issues of deadweight spending effects (Lenihan and Hart 2006).
As our results arise from a sample of firms that are believed to engage in R&D, as identified by the Irish Central Statistics Office (CSO), caution should be exercised when extrapolating the results of our study. This is a limitation of the current study, as such firms may have higher levels of internal capabilities for R&I vis-à-vis less innovative firms. Thus, it would be interesting if future research were to apply a similar analysis in other settings, and with more general and representative samples, to elucidate the extent to which the findings of the current study can be generalised to less innovative firms. Future studies may also usefully employ more detailed data on firms' access and use of external finance, in addition to firms' levels of leverage. Our findings, however, highlight that a more encompassing theory of the impact of firms' internal financial resources on firms' engagement in research and innovation might beneficially consider heterogeneities across different research and innovation activities, and firms of different sizes. Particularly, it may reconsider the extent to which financial resources trigger research and innovation in small-sized firms.

Scientific Research
Dev.

Scientific Research
Dev.

Scientific Research
Dev.

Scientific Research
Dev.

Scientific Research
Dev.

Scientific Research
Dev.

Scientific Research
Dev.

Org. Innovation
(1) (3) (1) (3)  Results presented in log likelihoods: Robust standard errors in parentheses. One-digit NACE Rev.2 codes and survey wave control variables included. Rho (2 to 7) refers to the correlation of the panels' residuals for each pairwise comparison. Statistically significant Rho coefficients support the simultaneous estimations of the probit models by means of a multivariate approach. *** p < 0.01, ** p < 0.05, * p < 0.1.   Results in average marginal effects: Robust standard errors in parentheses. One-digit NACE Rev.2 codes and survey wave control variables included. Rho refers to interclass-correlation, which is the variance that can be explained by differences across panels *** p < 0.01, ** p < 0.05, * p < 0.1.