This paper investigates the moderating role of firm age and size on the relationship between slack time and product innovation. The survey asks participating firms if firms introduced new or significantly improved goods and services in previous years for product innovation. The variable is measured as a dummy, which takes ‘1’ if the firm has introduced or significantly improved on it and ‘0’ otherwise. However, the innovation considered in this study complies with the Community Innovation Survey (CIS) and the Oslo Manual, where firms document if in the last three years they introduce any new or significantly improved forms of innovation. A second dependent variable is measured as the percentage of the main innovative products and services. The variable is used to perform the robustness checks in the analysis. The same variable is used to determine the moderating role of age and size in the relationship between slack time and product innovation for large firms and Small and Medium Enterprises (SMEs).

The measure of innovation has been employed by scholars in some studies (Ayyagari, Demirgüç-Kunt, & Maksimovic, 2011; Barasa, Knoben, Vermeulen, Kimuyu, & Kinyanjui, 2017; Chadee & Roxas, 2013; van Uden et al., 2017). Product innovation gives firms a competitive advantage and offers them a better position in the market by launching a better quality and product with cost reduction. This creates an impact which makes the firms to fill the gaps in demand (Galindo & Méndez, 2014) and to increase their market share (Leskovar-Spacapan & Bastic, 2007).

Firms that export are more efficient than non-exporting firms (Cleride et al., 1998). A few scholars have also documented the relevance of exporting capabilities to the performance of the firms (Ganotakis & Love, 2011, 2012; Filatotchev et al., 2009). The WBES reports the percentage of sales accounting for direct exports. This measurement is retained in this analysis as documented in the WBES.

The study controls for both private and foreign ownership because it could impact the ability of the firms to allocate slack time to generate innovative ideas. Ownership heterogeneity has been considered essential components of firm-level innovativeness (Chen, Li, Shapiro, & Zhang, 2014; Choi, Park, & Hong, 2012; Song, Wei, & Wang, 2015). Dachs and Peters (2014) provide evidence of the role of both foreign and domestic ownership on the likelihood of innovative-active firm contributing to employment growth. Also, foreign ownership has been evidenced to impact firm-level employment growth (Lipsey, Sjöholm, & Sun, 2013). These variables are measured as a percentage owned by private domestic firms and foreign domestic firms, respectively. The measurements are retained as reported in the WBES.

Capacity utilisation has been considered an essential element in the success of the firms and their production efficiency (Levy, 2007; Mazumdar & Mazaheri, 2005; Mukwate Ngui-Muchai & Muchai Muniu, 2012). It is also documented that manufacturing firms in SSA on average use 50% of their capacity (Mazumdar & Mazaheri, 2005). Expanding their capacity and increasing their capacity utilisation is considered essential to production (Fevolden, 2015). Capacity relies upon the resources, such as buildings, machinery and labour firms have available. Capacity utilisation is the degree to which that capacity is being utilised. When the firm is making full utilisation of all its resources, it means the firm is operating at maximum capacity or 100% capacity utilisation. It is an essential piece of information in the process of lowering waste in the production process and stimulating innovativeness (Fevolden & Grønning, 2010).

The estimations allow for country dummies. In the ES, there are three sector classifications: manufacturing, retail, and services firms. Sector specificity also plays an essential role in innovation performance, being mindful of endogeneity and sector-specific differences in innovation output (Brouwer & Kleinknecht, 1996). Considering the importance of the three sectors, and as argued by O'Sullivan (2006) that sectoral differences in the innovative activity do have prominent implications for the allocation of resources. We generate three-sector dummies: manufacturing, services, and retail, respectively, where retail serves as the base category in the estimations. On the other hand, we give room for the industry effect by entering industry dummies in all the estimations to rule out the likely impact of unobserved industry heterogeneities.

To test our hypotheses, we use the IV binary treatment model because of the binary nature of our instrument and endogenous variables. It is useful as an estimate for binary treatment models with a heterogeneous response to treatment both for observable and unobservable selection. The use of a binary treatment model in this paper depends on the work of Cerulli (2012). The binary treatment model with heterogeneous treatment to response aids to address the likely endogeneity issues that may evolve. The study employs the probit-2sls approach as one of the ingrained models in the instrumental variable treatment model. This approach is considered to be the most efficient amongst the three embedded approaches in addressing the relationship in which this study tries to predict. The reason of being the most efficient stems from the fact that it produces the smallest projection errors, and it addresses the issue of the binary estimators in the analysis, which include the chosen instruments (patent, and the extent to which political instability and corruption are of significant constraints to firms’ operations), the dependent and independent variables (Cerulli, 2012). Eqs. (1)–(3) are the econometric set-ups while the functional forms (4) and (5) represent the estimation equations, respectively.

We utilise the following general functional framework to examine the working hypotheses:

In the specification above, subscripts denote that these measures are across firms (i). The estimation approach addresses two probable drawbacks in our dataset. First, competitive impacts or other shocks in one year could have constant consequences, pointing to the likelihood of firm-specific autocorrelation over time. Second, with wide-ranging data, there is continuously the probability of provisional heteroskedasticity. While the Wooldridge test for autocorrelation (Wooldridge, 2003, 2011) is employed for the instrumental variable estimation. The result dispels the presence of multicollinearity in the estimations (Drukker, 2018; O’brien, 2007). The Breusch-Pagan test (uncentered VIF) estimations are below 10; a result that remains within the benchmark of both VIF and uncentered VIF.

Although other estimation methods are employed, the main estimates are performed using a binary instrumental variable treatment model (ivtreatreg) with a Probit-2sls option. We include controls for time so that the findings are robust to measurement errors that could be created by macroeconomic effects common to all firms in sub-Saharan Africa. There are two additional possible endogeneity problems that the estimation considers. First, firms may have persistently different levels of ST and diversity that is not appropriately managed. The inclusion of firm-specific controls has helped address the situation. Second, it is possible that firm-specific shocks, like an unpredictably high demand, which may permit a firm to instantaneously build outlays in ST and expand product lines, ensuing in more category.

On the other hand, a firm that forestalls a demand for superior variety, as a result of disparate matter, could create more substantial investments in ST in expectation of this effect. Either of these issues could lead to a correlation between ST level and the error term, leading to biased estimates. By tackling this kind of endogeneity, the ST level is treated as endogenous, and an instrumental variables estimation applied. The instrument group is selected about a previous study (Brynjolfsson & Hitt, 2003). It captures a firm's distinctive capability to react to surges in the crave for ST. Since the ideal estimator does not comprise an instrumental variables deviation, the estimation employs a probit-2sls approach that is embedded within the binary instrumental variable model. The IV estimation uses three instruments that conform with the assumption of having valid instruments. These instruments are patents, political instability and corruption.

This section begins with the report of the descriptive statistics and correlation matrix of the estimation variables that are reported in Table A2. With the sectoral, on average distribution, 41% of the firms belong to manufacturing, 24% retail, and 35% services. 61% are small firms, 27 medium-size firms, and 10% large firm. On average, 7.3% is related to the firms’ main innovative products and services. 54% of firms implement product innovation. Relatively, firms in sub-Saharan Africa engage in the implementation of innovation. While 18% represents the top managers with managerial experience, 28% of the staff are trained, and 53.2% of the workforce possesses a high school certificate; a measure the analysis uses as a human capital variable. On average, 42% of firms in the sample allot employees slack time to generate innovative ideas in the workplace with 75% of the firms reporting for product innovation. It, however, reveals that innovative firms in sub-Saharan Africa attach a higher premium to new idea generation to enhance innovation performance even though the practical applicability has not been reported as compared to the multinationals like Google, Twitter, and others. Relatively 11% of the firms conduct internal R&D and only 4% for external R&D. On average, 22% represent the capacity utilisation of the establishments. Percentage of sale from direct export stands at 4.4%. The shares of private domestic ownership and foreign domestic ownership are 73.3% and 9.8%, respectively. The following variables: patent, corruption and political instability serve as the instruments for the estimation in Table 2 where we employ an instrumental variable treatment model with a heterogeneous treatment to response (i.e. probit-2sls option). The instruments uphold the underlying assumptions underlying the use of valid and reliable instruments in the instrumental variable framework. 5% of the firms report having applied for a patent. 19% of the establishments report that political instability constitutes a major obstacle to their operation, and 25% consider corruption as a major obstacle.

The study originally examines the slack-innovation-performance relationship by moderating firm age and size with slack time to predict the implementation of product innovation. Four different estimations are performed with two distinct econometrics models owing to two distinct operationalisations of the dependent variables. While in the main estimations, we employ a probit-2sls estimation method owing to the binary nature of the dependent variable, the robustness estimation is performed with a Tobit model because of the censored dependent variable. We also perform further estimation using the Probit model to observe the dynamics and divergence in the results relating to the ivtreatreg with the 2sls output. With the Probit and Tobit estimations, the standard errors are adjusted for 282 clusters in country and industry, respectively. Table A1 reports the probit estimation, while Table A3 reports separate estimations for SMEs and large firms.

Now to the central estimation, this study hypothesises a positive relationship between firm age, firm size, slack time and innovation. The findings of Model 1, Table 2, in which all other variables are introduced in the estimation without the interaction variables, show that firm size (b=0.023, p<0.01) and slack time (b=0.228, p<0.01) support the introduction of product innovation. The results support our H1 and H3, respectively. However, there is no evidence for firm age supporting the introduction of product innovation; hence, our H1 is not confirmed. These results add to the inconclusive arguments of the role of age in the ability of firms to innovate. While the result upholds the findings of previous studies, it also contradicts the discoveries of related studies regarding the age-innovation performance relationship. However, other variables included in this estimation show a significantly positive relationship with product innovation except for managerial experience and external R&D where we observe a significantly negative relationship, and no significant relation for foreign ownership. The results follow in this order internal R&D (b=0.346, p<0.01); external R&D (b=−140, p<0.10); trained staff (b=0.093, p<0.01); managerial experience (b=−0.026, p<0.10); staff education (b=−0.0004, p<0.01); export (b=0.001, p<0.01); capacity utilisation (b=0.001, p<0.01); private ownership (b=0.0004, p<0.05).

Moreover, Model 2 in Table 2 begins with the interaction variables. The fourth proposition suggests that firm age and slack time could reinforce each other to support the propensity to which firms innovate. In this approach, we follow one of the steps proposed by Aiken et al. (1991) and Cohen et al. (2013) by introducing the interaction variable after the control and observing the idiosyncratic effects. The H4 is supported, in that, firm age significantly and positively moderates slack time to support the introduction of product innovation (b=0.016, p<0.10). Similarly, we implement the same pattern to observe the direct moderation effect between firm size and slack time, concealing the moderation of age and slack time in Model 3, Table 2. The finding is also compelling in that firm size and slack time (b=0.018, p<0.01) reinforce each other to support the introduction of product innovation. H5 is then supported. This indicates that both age and size efficiently moderate the relationship between slack time and firm-level innovation performance. In Model 4 still Table 2, we propose a three-way relationship, that is, age-size-slack time effect on product innovation. The result yields no significant return to innovation. This indicates that firm age, size and slack time could not combine efficiently to predict a return to the introduction of product innovation, indicating that H6 is not supported.

Furthermore, in Model 5, we implement a multilevel estimation by incorporating all the moderators. While we observe a substantial increase in the coefficients of slack time×age (b=0.046, p<0.01) and slack time×size (b=0.030, p<0.05); the three-way moderation turns out to be significant but with a decline to innovation (b=−0.013, p<0.01). However, in Model 6, we estimate the quadratic relations by getting the squared terms of both firm age and size and further moderate with slack time to predict a performance relationship. The idea in this framework is simply to observe predictive outcomes when the age and size of the firms are doubled. The results that emerge indicate that the linear quadratic relationships between size and product innovation are observed to be positive and significant (b=0.0044, p<0.01) while that of age is significant but negative (b=−0.0018, p<0.01). Besides, we perform a moderating relation of age2 and size2 with slack time to predict a performance return to innovation. While we do not observe any relationship between the moderating effect of size2 and slack time on product innovation, the result that emerges for age2 and slack time is positive and significant (b=0.0045, p<0.01). While the linear moderation of age, size and slack time is not significant (see, Model 4, Table 2), in the quadratic-linear estimation, we observe age2×size2×slack time to be significant but negative, although with a somewhat negligible coefficient (b=−0.0003, p<0.01). The relationship of the combined effects of squared terms of firm age, size and ST produces an inverted U-shaped. The same development plays out in the Probit estimation with a trifling divergence.

Thus far, the element of analysis has been a dummy variable capturing product innovation implementation. As the first robustness check, we re-examine our measurements (with some alterations) using a percentage of the main innovative products and services as a dependent variable in the framework. To do so, we employ a Tobit estimation maintaining the same collection of variables as employed in Eq. (4). The dependent variable is censored from the left, which justifies the utilisation of a Tobit model in establishing the association between the predicted and the explanatory variables.

The results that emerge from the robustness estimation diverge significantly from the results obtained in the central estimation. While we observe age to relate negatively with product innovation in Model 1, Table 3 (b=−3.11, p<0.01), no significant return obtained for firm size and innovation. Slack time is observed to be consistently significant across the entire model except in Model 5, that is the multilevel analysis. Similarly, while age moderates the relationship between slack time and innovation performance but inversely (b=−0.73, p<0.01), firm size significantly moderates the relationship between slack time and innovation (b=0.94, p<0.01). The three-way moderation also significantly but inversely predicts the slack-innovation relationship (b=−1.91, p<0.01). The quadratic estimation tends to differ marginally from the results in the central estimation. Both the quadratic coefficients of age (b=−0.35, p<0.01) and size (b=−0.30, p<0.01) are observed to relate inversely significant to the implementation of product innovation. Interestingly, the quadratic-by linear relations between slack×age (b=0.17, p<0.01) and slack time×size (b=0.74, p<0.01) are observed to be positive and significant. The three-way quadratic estimation as obtained in the primary estimation also predicts an inverted U-shaped relationship. All the control variables maintain the same levels of significance as with the case in the central estimation.

In addition to the above robustness check, utilising the same set of variables, and using the percentage of the main innovative product as a dependent variable, we further investigate the effect of the paired variables on both SMEs and large firms. Interestingly, the results that emerge in Table A3 relates more significantly positive to SMEs than large firms. Taking into account the first paired variables, ST×Size is observed thus: SMEs (b=5.89, p<0.01); ST×Age (b=2.72, p<0.01) and ST×Age×Size (b=−2.14, p<0.01). For large firms, we can only observe the results for ST×Size (b=11.8, p<0.05). Now Shifting to the quadratic estimations, the findings relate thus: for SMEs, ST×Size2 (b=0.59, p<0.01); ST×Age2 (b=0.22, p<0.01) and ST×Age2×Size2 (b=−0.065, p<0.01). For large firms, no significant result emerges. All the inclusive variables are found to be statistically significant for SMEs. An instance is the internal and external R&D. Both activities are found to be consistently significant and positive for all firm types. Indeed, while the definition of large firms and SMEs could vary regarding countries and economy size, these results reveal that if the top managers can manage the allocation of slack time more efficiently, its return to firm performance regarding product innovation is guaranteed.

The third aspect of the robustness check involves marginal effects. These graphs are captured in Fig. 3. We generate predictive marginal graphs for both the direct and quadratic interactions. The results that emerge in the plots reveal a further nuanced slack-performance relationship. Fig. B1 in the top left reveal a linear relationship between slack time and product innovation at the middle-size point of the interaction graph. At this point, the ability of a potential implementation of product innovation by firms is assured. There is also a latitude to which the movement of the line could either relate to a decline or an increase to product innovation. The graph shows that size plays a role in the offering of slack time to employees to generate innovative ideas. As envisaged, Fig. B2 at the top right graph is the quadratic state of firm size and how it moderates slack time to support the implementation of product innovation. It is observed that as the firm grows in size, the ability to implement the introduction of product innovation looks positive.

Fig. 3.

Predictive interactive marginal graphs.

(0.48MB).

Further, Figs. B3 and B4 show the slack-time-age interaction. The result emerging from Fig. B3 shows that the ability of firms to obtain returns to the implementation of product innovation when alloting slack time to employees to generate innovative ideas in the workplace is confirmed to be more productive when firms are younger. This reignites the ongoing debate regarding the ability of new entrants into the market to outsmart incumbents in the introduction of innovation. The situation, however, refutes the liability of newness hypothesis. While Fig. B4 reverses the observed case in Fig. B3, it shows that when firms grow older, the ability to generate returns to innovation with the utilisation of slack time is further illustrated. Both graphs reveal somewhat interesting extreme cases. While in the early stage of operations, new firms leverage their new products and capabilities to influence the market. At the latter stage of the firms, however, the stability of firms in the market could relate to knowing the dynamics in the market and probably have established a clientele-supplier relationship which allows them to leverage on employees’ slack time for innovation optimisation. We observe interesting dynamics when the initial size of the firms is considered (i.e. the number of employees when operations began). The initial size also moderates the relationship between slack time in supporting the implementation of product innovation. The result indicates that at the upper level of employees growth, the use of slack time remains supportive of firms innovative capabilities. The lower interactive graphs illustrate the three-way moderation of age, size and slack time. A nuanced picture also emerges on the relationship between firm age-size-slack-time in predicting the return to firm-level innovation.

The implications of our findings for practitioners and decision-makers are twofold. Although we contend that firm age and size can help firms transform employees’ slack time into performance more efficiently; practitioners and policymakers should be conscious that the age and size of the firms do not continually generate higher performance. As observed in the predictive margins for firm size, the midpoint (Fig. B1) reflects the efficient level that slack time efficiently support the implementation of product innovation. The squared function of firm size (Fig. B2) reveals an inverted-U shaped relationship between slack and innovation. Taking the age of the firm, the predictive margin graph (Fig. B3) relates that new entrant firms into the market could benefit from the utilisation of slack time to support innovation.

Similarly, the reverse tends to be the case when we square the age of the firm. It reveals that older firms could also benefit from the utilisation of slack time to complement the introduction of innovation (see, Fig. B4). This could be the case for small companies that have limited additional slack resources – for these firms, improvised rather than planned activities may be more conducive to higher performance. As the allotment of employees’ slack increases, the effects of managerial monitoring could increase as well to ensure a balance. Firms with abundant slack resources could exploit either their age or size to offset the negative effects of excessive slack and allot resources to attain improved performance.

Furthermore, an important revealed practical implication of the overall findings indicates that firms would gain from allotting slack time to employees because innovation flourishes on redundancy. If employees are constantly strained or stretched and have their minds always absorbed by the workplace bureaucracy, they are much less likely to think creatively compared to when they have some free time to think freely. Mostly, firms would have to extemporised from their fallowed resources to optimise performance. This study practically reveals that the logic of successful innovation is to amass capabilities and devise an appropriate means to use them for the benefit of the firms. Hence, top managers must be willing to cut the mustard when it concerns strategy deployment for performance. However, the fundamental predicament managers at times grapple with is whether, first, they should build up in surplus resources to cushion their firms from internal and external shocks and to hound novel prospects or whether, second, they should develop sinewy enterprises. Managers who are willing to go for broke would prefer to create both surplus slack resources, influential and strong enterprises at the same time. As a result, this study establishes that extemporised resource like the slack time, that might be daunting for top managers to allot effectively amongst the employees, contributes to firm performance; especially in the case of a pooled regression and SMEs. While the management of different slack resources mainly relies on the qualities of such resources and its perceived relevance to the success of a firm; this study finds that the firm-level attributes, age and size, are also quintessential to whichever strategy firms adopt to optimise performance.

Another essential area that offers a pragmatic implication concerns the perceived performance-effect of slack time for SMEs. There is a wind of caution that should be exercised by top managers of SMEs regarding the utilisation of slack time as a performance indicator. In the earlier review of the anecdotal effect of slack time, it shows that large firms have benefitted from the offering of slack time to their employees for creative ideas to support either the projects of the focal firms or the personal development of the employees themselves. However, if financial slack is considered more closely, it might still be more conducive for large firms to offer a higher amount of slack time to their employees as compared to SMEs. This financial slack (it refers to extra financial flows available to firms but not primarily part of the working capital) could help cushion the negative effect that might arise from the excessive usage of this extemporised resource, slack time, for performance. It then informs managers of SMEs that in case the ability to fall back on an extra financial resource is not available, this study suggests that top managers would have to prioritise the offering of slack time to desired or selected employees that are more connected to the overall performance goal of the firm. What this study has not tested is the role slack time could play for the R&D department or the overall R&D activities of the focal firms. In that regards, SMEs could try to narrow the allotment of slack time to the employees that constitute the R&D section, and through this strategy, they can limit the damage arising from the allotment and utilisation of slack time.

This study has some shortcomings that should be considered when interpreting the findings, and that also propose avenues for potential research in this area. First, the study sample consists of companies in sub-Saharan Africa that are farther from technological frontiers and whose operations are impeded by the weak institutions and leapfrogging economies. In that, the use of slack time might reveal a quite substantial divergence when compared with other regions of the world that are much closer to technological frontiers and boast of the quality of institutions. A generalisation of the findings obtained within the purview of firms that operate in Africa and other regions of the world that exhibit similarly political and economic configurations could hold. However, such may not constrain any generalizability of the findings to other countries with better institutional configuration.

From the data viewpoint, the study utilises cross-section data, and hence, the estimations could only confirm the link between the explained and explanatory variables within the framework of the dataset. Consequently, we could not assess the long-term impact of employees’ slack time on product innovation more robustly. Nevertheless, this does not emasculate the strength and relevance of our findings. Further, the issue of causation is also complicated while employing a cross-section dataset; so future research may address the possible causality concerns if the appropriate data is available. We espouse the idea of Atuahene-Gima (2005) that slack mirrors firms’ resources and market influence to utilise subsisting competencies, create new ones, and implement innovations, which is a rational and effortlessly useful representation of employees’ slack.

Slack has frolicked an essential role in organisational concepts of survival, growth, and performance. However, the argument still subsists regarding the optimum amount of slack to enhance firm performance. In this study, this argument is broadened by considering how firm age and size could play a role in the allotment of slack time to employees for innovative performance optimisation. We examine employees’ slack time and validate that its relationship with firm-level innovation performance is moderated by firm age and size. The results indicate that in earlier and advanced levels of firm age, there is a linear relationship between employees’ slack and firm performance, but in the midpoint of firm size and the point where firms document excess workforce, there is an inverse U-shaped relationship between slack time and innovation.

We hope these findings provide some basis for future studies to address the complex interaction of slack and other elements in the assessment of firm performance. A few intellectuals have indicated that management prudence given by resources and innovative tendency could also mediate the link between slack and performance (Bradley, Shepherd, et al., 2011; Bradley, Wiklund, et al., 2011; Cheng, Chang, & Li, 2013). Interested scholars could consider this avenue. Another potential area could involve the application of this scope to the other three innovation types to see how the dynamics of the results could play out. Also, the fact that most African countries do not diverge in some key institutional and economic factors, a comparative cross-country study could still be explored by considering a different time-period. An interesting research question that further emerges from this study concerns how diverse external knowledge sources moderate slack time to support firm-level innovation performance. A further area of study that emanates is the idea of a combination of strategy for performance optimisation. A complementarity or substitutability of strategy should be worthwhile in internal and external R&D and slack time; managerial experience and slack time;R&D staff and slack time. Also, another vital area for future research is whether employees’ slack time could serve as a substitute for external cooperation or help bridge the gap in knowledge divergence between firms. It could help inform top managers on the allocation of slack time to employees for the generation of novel ideas in the workplace as a means of optimising innovation performance.

Uncited references: This section comprises references that occur in the reference list but not in the body of the text. Please cite each reference in the text or, alternatively, delete it.

Academy and Oct (2007), Agarwal, Audretsch, and Sarkar (2014), Armstrong and Shimizu (2007), Arrow (1972), Aydın Unal, Unal, and Isık (2017), Prahalad and Hamid (2007), Terry Mousa and Chowdhury (2014), Valaei and Rezaei (2016), and Yoon, Lee, Lin, and Eyring (1976).