Recent studies demonstrate a robust, positive relationship between FCF and strategic investments in both advertising and R&D, with resource-rich firms most able to allocate substantial funds towards marketing and innovation activities (Ali et al., 2023; Sveikauskas et al., 2024). Empirical analyses confirm that the ‘affordability’ principle remains highly relevant: when internal financial resources are ample, firms reliably increase their advertising expenditures (Sveikauskas et al., 2024; Wu et al., 2022). Dynamic models further reveal that both R&D and advertising play mediating roles between FCF and sales growth, highlighting the strategic importance of resource allocation for firm performance (Ali et al., 2023).

Free cash flow is a key determinant of the firm’s R&D and advertising spending. Longitudinal evidence shows that rising FCF boosts both R&D and advertising expenditure, while declining FCF suppresses these investments. Thus, we hypothesise:

Generally, FCF might be more prevalent in technology companies. Industries in the high-tech sector – characterised by rapid technological change and heavy investment in R&D – rely extensively on FCF to fund large-scale innovation and sustain competitiveness (Liu & Zhang, 2023; Xin et al., 2019). Investing in R&D is critical for firms in uncertain and technologically complex industries, as these firms pursue new R&D initiatives to manage uncertainty, enhance strategic resilience, and leverage information about resource or market quality (Lin et al., 2021; Nguyen & Nguyen, 2024). Recent research confirms that basic and innovative R&D investments yield greater productivity gains in high-tech industries, with Chief Executive Officers (CEOs) in these sectors more inclined to allocate rising FCF towards high-risk R&D initiatives when resources are available (Nguyen & Nguyen, 2024; Xin et al., 2019). We hypothesise:

The resource-based view posits that a firm achieves and sustains competitive advantage by possessing resources and capabilities that are valuable, rare, inimitable and non-substitutable (Barney, 1991). Within this theoretical framework, intangible resources such as R&D and advertising are recognised as essential for long-term performance, with their value stemming from their tacit and immobile nature. Recent empirical research further establishes that R&D and advertising function as strategic intangible assets that can drive various dimensions of firm performance, with R&D exerting a stronger influence on firm value and advertising enhancing profitability, even under conditions of environmental volatility (Piao & Choi, 2022).

Innovation remains fundamental to achieving and sustaining competitive advantage in both advanced and emerging economies (Giudici et al., 2017). Contemporary studies find that increased R&D expenditure is consistently linked to enhanced innovation outputs at both the firm and national levels (Ivanová & Žárská, 2023; Kahn et al., 2024). Furthermore, empirical analyses demonstrate that technological innovation, together with R&D investments, plays a crucial role in driving economic growth and strengthening scientific production across diverse contexts (Arana-Barbier, 2023).

Firms create and accumulate knowledge through their internal R&D activities, which not only spur firm-level innovation but also enhance productivity and knowledge spillovers to the broader economy (Audretsch & Belitski, 2023). Corporate R&D expenditure and patent accumulation represent vital measures of knowledge flows and stock within firms, both serving as key determinants of innovation capacity and competitive performance (Audretsch & Belitski, 2023; Mallinguh et al., 2022). The interplay between technological knowledge creation – spurred by sustained R&D – and complementary skills, including technology transfer and patent portfolio management, has significant implications for firm growth and competitive heterogeneity (Arana-Barbier, 2023; Mallinguh et al., 2022). Empirical evidence further demonstrates that firms with advanced innovation outputs and higher R&D intensity consistently report notable improvements in productivity, employment and sales, reflecting the enduring value of innovative investment for long-term performance (Kahn et al., 2024).

Advertising serves as a key channel for firms to communicate value, enhance sales, and strengthen firm-level profitability and competitive positioning (Piao & Choi, 2022). The signalling theory underscores how advertising signals product quality to consumers, thereby influencing purchasing decisions (Kihlstrom & Riordan, 1984). Recent evidence shows that advertising investments, particularly when complemented by digital marketing capabilities, provide significant profitability gains beyond those achieved by traditional marketing alone (Homburg & Wielgos, 2022). Resource-based logic positions advertising and R&D as complementary intangible assets, both of which enable differentiation and sustained firm performance (Piao & Choi, 2022). Innovation capability, when integrated with intellectual capital and dynamic marketing resources, forms a multidimensional basis for ongoing innovative performance (Ali et al., 2023). Contemporary perspectives suggest that brand advertising aims to deliver a meaningful, compelling and achievable promise to consumers, with innovation playing a dominant role in shaping this message (Martin et al., 2024). Altogether, R&D and advertising are strategic resources that enhance firm performance and support lasting competitive advantage (Piao & Choi, 2022). Grounded in resource-based and knowledge-based perspectives, the analysis posits the following temporal patterns: increases (or decreases) in FCF are expected to result in corresponding rises (or reductions) in R&D and advertising expenditures, subsequently influencing sales in the same direction. The mediating role of R&D expenditure change is predicted to be more substantial in high-tech industries than in other sectors. Since FCF change is positively correlated with changes in both R&D and advertising outlays, which in turn are associated with sales change, and industry type moderates the link between FCF change and R&D expenditure change, the hypotheses are as follows:

The hypothesised relationships among FCF change, R&D and advertising expenditure changes, industry type, and sales change are summarised in the conceptual framework diagram presented in Figure 1. The diagram reflects the dynamic mediation structure, moderating effect and relevant time periods for all focal variables included in the empirical model.

The sample firms used in this study encompass both high-tech and non-high-tech firms in North America. Longitudinal financial and organisational data were obtained from established sources, including Compustat, U.S. Securities and Exchange Commission (SEC) filings, the Thomson Reuters Institutional Holdings (13F), and the Institutional Brokers’ Estimate System (IBES), covering the fiscal years 2012–2019. This timeframe was deliberately chosen to focus on a period of relative macroeconomic stability, thereby excluding the effects of the global financial crisis, the European debt crisis before 2012 and the unprecedented structural disruptions posed by the COVID-19 pandemic after 2019.

Restricting the sample to this pre-pandemic interval ensures that observed changes in FCF, R&D and advertising expenditures reflect typical firm behaviour rather than extraordinary responses to exogenous shocks. This design strengthens the interpretability of dynamic mediation and longitudinal analyses, as confounding effects from crisis-driven policy interventions and demand volatility are systematically avoided. Furthermore, 5 consecutive years of data per firm are required to enable robust estimation of random slopes and the dynamic mediating and moderating effects.

Consistent with prior longitudinal management research, our choice of these years supports comparability with existing benchmarks and establishes a necessary baseline for future studies analysing post-pandemic change. Financial and utilities firms were excluded because of fundamentally different capital structures and regulatory regimes (Klasa et al., 2018). Data for focal and control variables cover 5 consecutive years for each firm. Out of an initial pool of 5388 firms, 4673 were excluded because of incomplete data on the primary variables of interest, leaving 715 firms with full longitudinal coverage. After removing an additional 15 firms with missing control variables, the final sample comprised 700 firms (193 high-tech and 507 non-high-tech, classified using Standard Industrial Classification (SIC) codes per U.S. SEC Office of Technology guidance), yielding 3500 firm-year observations (5 × 700) for hypothesis testing.

Sample selection bias can occur when a considerable amount of data is missing. Heckman’s two-step procedure (Heckman, 1976) can be employed to examine and correct for potential sample selection bias. A probit model is applied to the binary (missing or non-missing) data on focal variables for 2411 firms with complete data on the control variables.

To capture change over time, we use the slope of the linear growth trajectory, encompassing a 5-year span. How to obtain the slopes of the linear growth trajectories of the focal variables and specify the relationships among these slopes will be elucidated in the next section. Free cash flow is measured as ‘the cash flow from operating activities minus capital maintenance expenditures’ (Mills et al., 2002). Research and development and advertising are two types of investments in intangible assets for which expenditure data are available. Sales refer to net sales, calculated as gross sales less discounts and sales returns and allowances. Sales data are from 2015 to 2019. Both advertising and R&D expenditures exhibit time-lag effects on sales, with the former often generating more immediate impacts than the latter (Cohen et al., 2010). A 2-year lag is applied between R&D expenditures and sales (Chun et al., 2015). Therefore, R&D expenditure data span from 2013 to 2017. A 1-year lag is adopted for advertising expenditures (Kim & Joo, 2013), covering advertising expenditures from 2014 to 2018. Free cash flow also has time-lag effects on R&D and advertising expenditures, with faster effects necessitating a 1-year time-lag. Due to varying time lags, two FCF variables are used. The data for the first FCF variable span from 2012 to 2016, while the data for the second one cover the period from 2013 to 2017. Note that lagging offers another advantage: It can help alleviate concerns about endogeneity (Pitariu & Ployhart, 2010).

We control for the following firm attributes:

Capital-labour ratio. The capital-labour ratio (denoted by CLR) is the ratio of capital to total labour employed. Firms tend to have a higher CLR when they aim to improve labour productivity through capital investment.

Firm leverage. Firm leverage (denoted by LVRG) is related to firm performance (Burrus et al., 2018; Fosu, 2013). In this study, LVRG is assessed as the ratio of total debt to total equity (Burrus et al., 2018).

Financial constraints. Financial constraints are constraints that come with money. Financial constraints exert influence on firm expenditures and firm performance (Golovko et al., 2022). In this study, financial constraints are measured using the size-age (SA) index developed by Hadlock and Pierce (2010).

Corporate governance. Corporate governance mechanisms are widely studied for their influence on firm outcomes, with institutional ownership (IO) playing a significant monitoring role in shaping firm performance, particularly in emerging economies (Abedin et al., 2022). Following established empirical practice, this study adopts both IO and analyst following (AF) as proxies for corporate governance, in line with prior literature (Chung et al., 2003).

Firm size. Firm size (denoted by SIZE), a standard control variable, is measured here as the natural logarithm of total assets (e.g. Fosu, 2013).

Firm age. Firm age (denoted by AGE), also a common control variable, is calculated as the number of years between a firm’s establishment and the year 2019.

We use a two-step approach to analyse dynamic mediation. The first step uses growth modelling to capture change over time and provides estimates for unobserved random slopes of the linear growth trajectories of FCF, R&D expenditures, advertising expenditures and sales for all firms, considering time lags. In the second step, SEM is used, with the slope estimates obtained from the first step as input, to analyse the structural dynamic relationships, including the direct effects of prior FCF changes on subsequent sales change, indirect effects through subsequent R&D expenditure change and advertising expenditure change, and the moderating effect of industry type (high-tech vs. non-high-tech) on the indirect effect through R&D expenditure change.

We modelled the moderating effect of industry type (high-tech vs. non-high-tech) on the relationship between FCF and R&D expenditure only, as there is no theoretical basis to support the moderating effect of industry type on the FCF–advertising expenditure relationship. Resource-based and innovation perspectives suggest that R&D investment in high-technology sectors is particularly contingent upon the availability of resources. In contrast, advertising expenditures in both high-tech and non-high-tech firms tend to be more routine and exhibit limited industry-level variation in response to fluctuations in FCF.

Because time lags for the effects of FCF vary, we employ two FCF variables. Let SALES_it represent the sales of firm i at time t, t = 2015–2019, FCFLAG3_it–3 represent 3-year-earlier FCF (2012–2016), FCFLAG2_it–2 represent 2-year-earlier FCF (2013–2017), RD_it–2 represent 2-year-earlier R&D expenditures (2013–2017), and AD_it–1 represent 1-year earlier advertising expenditures (2014-2018). The level-1 growth model for FCFLAG3_it–3 is given by (Equation 1):

where i = 1, …, n (sample size, 700 in this study), t = 2015–2019, Time denotes the independent variable for the five time points (5 years), and ε_1it–3 are level-1 error series. Since the error series tends to be autocorrelated and autocorrelation should not be ignored, AR(1) (the first-order autoregressive process), commonly used for capturing autocorrelation (Littell et al., 2006, p. 175), is specified in the level-1 growth model. η_10i and η_11i denote the intercept and slope of the linear growth trajectory of FCFLAG3 for firm i. They are random coefficients and are specified in the level-2 growth model given by (Equation 2):

where γ₁₀ and γ₁₁ denote, respectively, the means of the random intercept η₁₀ and the random slope η₁₁, and δ_10i and δ_11i are level-2 errors for firm i. δ_10i and δ_11i are correlated errors with variances σδ102,σδ112 and covariance σ_δ10δ11. The growth models for RD_it–2, FCFLAG2_it–2, AD_it–1, and SALES_it are specified in the same way.

Growth modelling can be handled by using hierarchical linear models (e.g. Chen et al., 2011). Estimates of random coefficients for individuals can be obtained with empirical Bayes estimation, taking into account the information from the entire sample. The empirical Bayes estimates are more precise than those resulting from OLS (Chen et al., 2011).

In the second step of the approach, we analyse the dynamic relationships among the random slopes by using SEM with the empirical Bayes estimates obtained from the first step as input. In addition to the firm attributes mentioned in the previous section, the average levels of FCF, R&D expenditures and advertising expenditures across years were held constant (Chen et al., 2011). Let FCFLAG3change, RDchange, FCFLAG2change, ADchange and SALESchange represent, respectively, η_11, η_21, η₃₁ η₄₁, and η₅₁, denoting the random slopes (the rates of change) of the linear growth trajectories of FCFLAG3, RD, FCFLAG2, AD, and SALES. The assessment of the dynamic mediation requires Equation 3, Equation 4, Equation 5 and Equation 6, presented as:

where i = 1, …, 700, HT denotes the dummy variable with value of 1 if the firm is in a high-tech industry and 0 otherwise, FCFAVG, RDAVG, ADAVG, CLRAVG, LVRGAVG, SAAVG, IOAVG, AFAVG, and SIZEAVG denote, respectively, the average levels of FCF, R&D expenditures, advertising expenditures, CLR, LVRG, SA, IO, AF and SIZE across years, AGE denotes the firm age, and ν₁ − ν₄ are uncorrelated errors. β₁₁ denotes the effect of 3-year-earlier FCF change on 2-year-earlier RD change. ϕ₁₃ denotes the moderating effect of HT on the relationship between 3-year-earlier FCF change and 2-year-earlier RD change. Thus, ϕ₁₁ is the effect of 3-year-earlier FCF change on 2-year-earlier RD change for non-high-tech firms, and (ϕ₁₁ + ϕ₁₃) is that for high-tech firms. ϕ₂₁ denotes the effect of 2-year-earlier FCF change on 1-year-earlier AD change. ϕ₃₁, ϕ₃₂, ϕ₃₃ and ϕ₃₄ stand for, respectively, the effects of 3-year-earlier FCF change, 2-year-earlier RD change, 2-year-earlier FCF change, and 1-year-earlier AD change on sales change. It follows that the indirect effect (i.e. the mediating effect) of FCF change on sales change through RD change is given by β₁₁ϕ₃₂, that for high-tech firms is given by (ϕ₁₁ + ϕ₁₃)ϕ₃₂, and that for non-high-tech firms is given by ϕ₁₁ϕ₃₂. The difference in the indirect effects between high-tech and non-high-tech firms is ϕ₁₃ϕ₃₂. The indirect effect of FCF change on sales change through AD change is ϕ₂₁ϕ₃₄. Structural equation modelling can be used to test for these effects. Given that the data are non-normal, the Satorra and Bentler (1994) correction, often referred to as a robust maximum likelihood method, is utilised. The tests are right-tailed. The variables are all winsorised at the upper and lower 1% to mitigate the influence of outliers.

As previously mentioned, of the 2411 firms with complete data on the control variables – CLRAVG, LVRGAVG, SAAVG, IOAVG, AFAVG, SIZEAVG and AGE, only 700 firms have non-missing data for the focal variables. Heckman’s two-step procedure consists of the substantive equation (one of Equations [3–6]) and the selection equation (with the predictors of the control variables). If the disturbances associated with the two equations are correlated, sample selection bias exists. The probit model in Step 1 is used with the dummy endogenous variable coded as 1 when all focal variables’ data are present and 0 otherwise to model selection. The resulting new variable, known as the inverse Mills ratio, is then added to the substantive equation in Step 2 to correct for the bias. On the other hand, if the disturbances are uncorrelated, there exists no sample selection bias and no correction is needed. The task can be handled by using PROC QLIM in SAS/ETS (SAS Institute Inc., 2015a). In this study, there is no sample selection bias, as the disturbances associated with the two equations are small and not significantly correlated (p > 0.78).

Given the non-normality of the data, robustness was assessed using the bias-corrected bootstrap approach. Simulation evidence from MacKinnon et al. (2004) indicates that this method provides superior resampling accuracy overall. For the right-tailed test for an effect, we examine the lower confidence limit only. If the lower bootstrap confidence limit at the 100(1 – α)% confidence level is greater than zero, then the effect estimate is significantly positive at the α significance level.

This article followed all ethical standards for research without direct contact with human or animal subjects.

Our findings underscore the strategic importance of FCF allocation for driving innovation and market presence. Firms with significant increases in FCF should increasingly allocate resources to R&D and advertising expenditures. These investments offer firms a dual advantage of capturing innovation and augmenting market share, thereby fostering growth and competitiveness. Notably, our findings indicate that firms experiencing higher rates of FCF change can leverage these advantages by correspondingly enhancing their R&D and advertising expenditure change rates. Industry type emerges as a moderator, with high-tech firms benefiting more from increased R&D investments. The sales gains associated with greater FCF directed to R&D are especially strong in high-tech industries, highlighting the need for industry-informed resource allocation strategies. This research also affirms that achieving immediate benefits should never compromise the pursuit of sustainable, long-term value creation. Prudent investments in R&D and advertising, rather than immediate profit-generation activities, are crucial for sustaining competitive advantages.

Investing in its own R&D constitutes a significant opportunity for a firm to accumulate critical competencies through cumulative learning and knowledge accumulation, ultimately enhancing sustainable competitive advantages (Barley et al., 2017). Restraining R&D expenditures in order to sustain short-term financial gains undermines the creation of enduring value, as the long-term benefits of R&D spending are well-established while short-term effects are limited (Ravšelj & Aristovnik, 2020). The importance of FCF and R&D investments in value creation has been affirmed (Khidmat et al., 2019), suggesting the need for careful consideration before undertaking mergers, rewarding leadership or distributing to shareholders.

A striking example of the long-term consequences of misallocating FCF can be seen in Intel Corporation. Former Intel CEO Paul Otellini famously declined an offer from Apple’s Steve Jobs to develop chips for the first iPhone, ultimately ceding a lucrative market to competitors (Madrigal, 2013). Similarly, after CEO Brian Krzanich resigned in 2018, his successor, Bob Swan, prioritised stock buybacks and cost-cutting over R&D investment.

While these financial manoeuvres temporarily boosted stock prices, they drained funds that could have fuelled critical innovation. As a result, Intel lost its technological leadership to rivals like Advanced Micro Devices (AMD) and Taiwan Semiconductor Manufacturing Company Limited (TSMC), highlighting the risks of underinvesting in R&D for short-term financial gains. This reinforces the importance of maintaining a forward-looking approach to FCF allocation, ensuring that firms – especially those in high-tech industries – continue to invest in innovation rather than prioritising immediate shareholder returns at the expense of long-term competitiveness.

Effective coordination between R&D and marketing functions can create synergistic advantages that often drive organisational success (Griffin & Hauser, 1996). This type of integration may encompass areas such as facility planning, workforce movement, structural organisation, reward systems and managerial procedures. Joint initiatives between R&D and advertising further strengthen the positive impact on sales growth.

Our study expands prior research by further clarifying how R&D and advertising expenditures dynamically mediate the connection between FCF and sales. The dynamic mediation framework shows that investing FCF in R&D and advertising is beneficial for a firm’s sales performance. Notably, the impact of allocating FCF to R&D is found to be stronger in high-tech sectors than in other industries, highlighting a greater effect for such firms.

While increases in FCF are generally linked to greater sales, Brush et al. (2000) provide empirical support for the agency theory argument that in firms with abundant FCF and weak governance, managers may drive sales growth that lacks profitability. Strong governance is crucial, as it compels managers to invest FCF wisely.

Some useful practical recommendations are provided. Firstly, for firms with substantial changes in FCF, this study emphasises the importance of aligning resource allocation with strategic objectives. Specifically, as FCF increases, managers should allocate a proportionate share of the incremental resources to R&D and advertising. Doing so not only fosters innovation but also strengthens market presence, driving long-term growth and competitiveness. Secondly, industry type matters in resource allocation decisions. The dynamic mediating effect of R&D investments is more pronounced in high-tech industries. High-tech firms, characterised by technological volatility and innovation intensity, should be particularly attentive to the need for allocating increased FCF to R&D. Non-high-tech firms should consider a more balanced allocation between R&D and advertising to achieve better results. Thirdly, collaboration between R&D and advertising can lead to more innovative product development and more effective advertising strategies, creating synergistic effects that drive sales growth. Fostering an environment where these two functions work in tandem can yield substantial benefits. Fourthly, this study affirms that prioritising immediate returns should never undermine the pursuit of enduring value creation. While it may be tempting to channel FCF into immediate profit-generation activities, prudent investments in R&D and advertising can create enduring competitive advantages. Firms should avoid prioritising immediate financial returns at the expense of sustained investment in these essential areas. Lastly, given the dynamic nature of the relationships examined in this study, businesses should establish mechanisms for continuous monitoring and adaptation. Regularly assessing the effectiveness of resource allocation decisions and adjusting strategies in response to changes in FCF can help firms stay agile and competitive. By capitalising on the insights presented here, firms can make strategic decisions with greater precision and enhance their ability to sustain competitive advantages.

This research acknowledges certain limitations that highlight avenues for future investigation. Firstly, the temporal scope of the study is confined to the period from 2012 to 2019; further work is needed to determine if these results hold in other time frames. Secondly, the carryover effects of R&D and advertising are not taken into account. Thirdly, future research could expand beyond the binary high-tech vs. non-high-tech distinction. Use of more granular or alternative industry groupings (e.g. technology intensity, sectoral innovation levels or sub-industry clusters) might reveal additional or different moderation patterns, not only for R&D but also for advertising pathways. Fourthly, in addition to industry type (high-tech vs. non-high-tech), future research could examine other industry-level moderators that may shape the dynamic relationships among FCF, strategic investments and firm performance – such as industry growth rate, competitive intensity, technological dynamism, capital intensity, regulatory environment, environmental uncertainty, innovation or advertising intensity and industry life cycle stage. Examining these factors could provide deeper insights into the mechanisms through which the dynamic mediating effects identified in this study are moderated across different industry contexts. Fifthly, Roper and Hewitt-Dundas (2015) report that both R&D-driven and externally sourced knowledge flows yield positive impacts on innovative sales. Future studies could examine the dynamic mediating effect of external knowledge flows on the relationship between FCF change and sales change. Finally, the present analysis intentionally excludes the period affected by the COVID-19 pandemic in order to focus on dynamic patterns under relatively stable economic conditions. This decision was made to avoid confounding effects arising from extraordinary economic shocks and policy interventions present during the pandemic, which can substantially distort firm-level financial and strategic behaviours. However, we recognise that examining how the dynamic mediation pathways involving R&D and advertising expenditures may shift during the crisis is an important issue for understanding firm adaptation. Future research could address this limitation by extending the temporal window to include the pandemic years and comparing the mediating effects before and after this macroeconomic disruption. Such an extension would further illuminate how external shocks reshape resource allocation and performance dynamics.