Assessing the Impact of Capital Expenditure on Corporate Profitability in South Korea’s Electronics Industry: A Regression Analysis Approach

Abstract

This study investigates the relationship between capital expenditure (CAPEX) and long-term corporate profitability in South Korea’s electronics industry. Using panel data from 126 listed electronics firms covering 2005–2019, the research applies fixed-effects regression analysis to examine how CAPEX influences profitability, measured by EBITDA/total assets. The results confirm that CAPEX exerts a positive and statistically significant long-term effect on profitability, with stronger but not significantly different impacts for large firms compared to SMEs. The findings contribute to empirical evidence on capital investment efficiency and the implications of economies and diseconomies of scale in capital-intensive industries.

1. Introduction

Capital expenditures (CAPEX) are key to long-term company success and competitiveness. They play an important role in improving productivity and contributing to profit growth. Capital investments enable corporations to expand capacity and constantly develop and innovate manufacturing processes. However, capital expenditures do not have an immediate effect; rather, there are time lags between investment and the realisation of benefits [1]. In this sense, management’s decision-making regarding capital investments is crucial, considering the risks and rewards for internal and external stakeholders [2]. Although capital expenditure and profitability are closely related through capacity expansion, productivity gains and technological upgrading, there is no strict causal relationship between them. Instead, both variables are shaped by broader strategic, operational, and financial decisions made within the corporation. For this reason, the aim of this study is not to establish a deterministic causal link, but rather to examine the long-term association between CAPEX and profitability in a setting where investment cycles are structurally important. This distinction is important because investment outcomes usually take a long time to materialise, and their financial effects depend on market conditions, industry characteristics, and firm-specific capabilities. By focusing on long-term profitability, this study aligns its analytical approach with the inherently delayed nature of capital investments.

Considerable attention has been given to the influence of CAPEX on a firm’s performance based on empirical evidence. However, the literature shows mixed findings. Many researchers contend that there is a positive correlation between capital expenditures and firm performance, such as profitability, productivity, and value. Nevertheless, it has also been argued that capital expenditures negatively affect firm performance. Another view is that there is no statistical relationship between capital expenditures and firm performance [3,4,5,6]. Nonetheless, previous studies investigating the relationship between capital expenditures and firm performance have comparatively shed little light on firm size, which is treated as one of the predictors of corporate performance. Most researchers focus on large firms or firms included in key indexes. Additionally, the electronics industry in Korea, the fourth-largest electronics producer in the world starting in 2016 with 117 billion euros in revenue, has not received much attention in this field of study [7,8]. Therefore, this study aims to address this research gap by presenting empirical results on the effect of capital expenditures on the performance of electronics companies in Korea, offering insight into how this effect differs depending on firm size. This study examines the effect of capital expenditures on long-term profitability and how firm size affects this outcome. The following research question will be addressed: How does the magnitude of the impact of capital expenditures on long-term profitability differ according to firm size?

This study contributes to the existing literature by investigating how long-term capital expenditure influences differ depending on company size. Additionally, this study will enrich the empirical evidence in the field of capital expenditures by investigating the financial data of technology- and capital-intensive electronics companies in South Korea that require heavy investments in research and development (R&D) and production facilities.

The rest of this paper is organised as follows: Section 2 reviews and summarises the relevant existing literature, followed by hypothesis development. Section 3 presents sample data and research methodology. The results of the data analysis are presented and discussed in Section 4. The conclusion and limitations of this research are presented in Section 5.

2. Literature Review and Hypothesis Development

2.1. Capex and Financial Performance

There are six analytical perspectives through which financial performance can be assessed: market-value performance, profitability, efficiency, leverage, liquidity, and growth. Previous studies on capital expenditures have primarily focused on profitability, typically measured by ROA [5,9,10,11,12,13]. A second stream of studies links CAPEX to efficiency, often via fixed-asset turnover, while a third considers CAPEX as a driver of growth and examines its association with revenue expansion. While these three perspectives offer complementary insights, most studies focus narrowly on single indicators, thereby limiting the development of an integrated analytical framework through which CAPEX influences firm performance across multiple dimensions.

Regarding profitability, Singh et al. [12] demonstrated a positive relationship between capital intensity (CAPEX/revenue) and ROA using a cross-country sample of 120 large firms. While this global dataset strengthens external validity, using capital intensity rather than CAPEX flows introduces conceptual ambiguity because the ratio is also strongly affected by firm size. Likewise, Taipi and Ballkoci [13] reported a positive effect of CAPEX on ROA for 30 Albanian construction firms between 2008 and 2015; however, the narrow sectoral scope and small sample size limit generalizability. Amini et al. found that physical capital investments increased operating income the following year using data from 2000 of the world’s largest R&D investors [14]. However, this sample is biased toward highly innovative firms, whose investment dynamics differ from those of typical industrial corporations. Majanga [15] found that CAPEX positively correlates with ROCE and NPM in Malawi. However, frontier-market conditions, such as reduced liquidity, market imperfections, and higher macroeconomic volatility, may limit the applicability of these findings to advanced economies. Curtis et al. observed that current CAPEX positively affects net income over a five-year horizon among U.S. firms from 1980 to 2017 [16]. However, structural shifts in accounting standards, technology cycles, and financial markets over nearly four decades introduce comparability challenges. Kim et al. [17] further illustrates the heterogeneity of CAPEX effects by showing that CAPEX lowers short-term performance in loss-making firms but improves it in profitable firms. However, their reliance on market-based profitability (net income/market value) makes the findings sensitive to fluctuations in investor sentiment and short-term valuation noise.

Contradictory evidence comes from Dovita et al., who found no relationship between CAPEX and ROA among 35 Indonesian consumer goods firms from 2014 to 2017 [5]. Similarly, Udoayang et al. [1] reported no significant effect of CAPEX on profit after tax or fixed-asset turnover for 69 Nigerian firms (2013–2018), though they identified a positive association with operating income. These mixed results underscore that the relationship between CAPEX and profitability depends on firm characteristics, capital market development, and measurement choices. These limitations challenge the comparability of cross-study findings.

A smaller but theoretically important body of literature compares the short- and long-term impacts of CAPEX. Turner and Hesford examined 305 hotel renovation projects (2004–2010) and found that CAPEX enhances short-term performance but depresses long-term metrics, such as revenue per available room [6]. However, their industry-specific context, characterised by cyclical demand and asset-heavy operations, may limit its applicability to non-service sectors. Conversely, Kwistianus and Juniarti showed that CAPEX improves the persistence of ROA and ROE in large-cap Indonesian firms (2016–2019) but has no short-term effect, supporting the view that financial returns from capital investments take time to materialise [9]. Together, these studies highlight the strategic horizon problem: CAPEX may produce delayed, nonlinear, or even contradictory effects depending on the analysed time window.

Although research and development (R&D) is conceptually distinct from CAPEX, studies of R&D investment offer valuable parallels because both represent long-term resource commitments and intangible or tangible capital accumulation. Nandy [10] documented the positive effects of R&D intensity on profitability and market value among Indian pharmaceutical firms. Amini et al. found that R&D expenditures increase subsequent operating income [14], and Curtis et al. observed positive effects on net income over five years [18]. Ahn et al. found that SMEs and large firms with rising R&D investments achieve higher revenue growth [19]. Lee and Lee [20] further demonstrated that the relationship between R&D intensity and Tobin’s Q becomes negative under high leverage. This underscores the moderate role of financial constraints. Park et al. [21] confirmed the positive association between R&D investment and revenue growth using a global sample. Collectively, these findings demonstrate that long-term investments, whether tangible or intangible, often influence performance through extended temporal lags and are shaped by financing conditions.

Another strand of research investigates investment spikes, which are exceptionally large increases in tangible asset investment. Yu et al. found that investment spikes increase short-term revenue in Chinese manufacturing firms but suppress long-term revenue growth, suggesting saturation or overcapacity effects [22]. In contrast, Gradzewicz [23] observed that investment spikes generate immediate and persistent revenue growth among Polish companies. These conflicting results highlight methodological challenges, including baseline selection for identifying “spikes,” sectoral differences in capital intensity, and variations in demand elasticity.

In summary, empirical findings on the relationship between capital expenditure and financial performance are mixed, reflecting significant heterogeneity in settings, methodologies, performance metrics, and investment horizons. While many studies document positive profitability effects, others reveal negative or insignificant relationships. Nevertheless, the literature converges on two essential principles: first, CAPEX enhances a firm’s productive capacity, and second, the financial impact of CAPEX unfolds over time rather than immediately. These insights underpin the long-term nature of CAPEX and motivate the first hypothesis of this study:

Hypothesis 1.

Capital expenditure has a positive impact on a corporate’s long-term profitability.

2.2. Firm Size and Financial Performance

Firm size influences financial performance by affecting scale economies, capital absorption capability, access to financing, and managerial and technological resources. According to OECD evidence [24,25], large firms, particularly in manufacturing, demonstrate higher productivity and scale efficiency, leading to greater profitability potential.

However, empirical studies on firm size and profitability alone show mixed results, largely due to differences in measurement. Some studies report that firm size, measured by total sales or total assets, improves ROA and ROE [26,27,28]. Others document negative or insignificant relationships, particularly when using total assets [29,30,31]. These inconsistencies suggest that firm size alone is not a reliable predictor of profitability. Rather, the effectiveness of size depends on how well firms deploy capital.

Studies using market-based performance metrics also report inconsistent results. Some studies find that total assets enhance firm value [32,33,34], while others find no relationship [35,36,37] or a negative effect during periods of heightened uncertainty [38].

Despite the mixed empirical patterns, the theoretical and structural logic is clearer. Large firms have a greater capacity to absorb, implement, and leverage capital expenditures due to their superior scale, diversified revenue bases, stronger internal financing, and more robust complementary assets, such as technology, managerial expertise, and distribution networks. By contrast, SMEs face financing constraints, higher capital costs, and limited risk tolerance. They also often lack the operational depth required to exploit large CAPEX commitments. Thus, given that large companies benefit more from economies of scale than small- and medium-sized companies when manufacturing up to the optimum production level, the second hypothesis of this research is as follows:

Hypothesis 2.

Large-sized corporations show a stronger positive impact of capital expenditure on their long-term profitability than small- and medium-sized corporations.

2.3. Combined Theoretical Model

Bringing these streams together yields an integrated model consisting of three core mechanisms.

Mechanism 1: Direct CAPEX effect

CAPEX enhances productive capacity. After an adjustment and utilisation period, this translates into higher long-term profitability (supported by [13,14,15,18,39]).

Mechanism 2: Firm size as a capability factor

Large firms have structural advantages, such as economies of scale, financing strength, and asset diversification, that improve capital efficiency and mitigate the risks of long-term investment [25,26,27,28,40].

Mechanism 3: Moderation of the CAPEX–profitability link

The ability of CAPEX to generate long-term profitability varies by firm size. Large firms can implement capital projects more efficiently due to scale and complementary assets. SMEs face financial constraints and higher investment risk, which weakens CAPEX realisation. Empirical inconsistencies in size–profitability studies do not undermine the importance of analysing size as a moderate factor but rather emphasise it.

Thus, the model predicts that the long-term value created by CAPEX is magnified in large firms and attenuated in smaller ones.

3. Data Collection and Methodology

3.1. Data Collection

As of 21 May 2025, a total of 290 firms in the electronic components, computer, video, audio, and communication equipment manufacturing industry were listed on the Korean stock market. These firms are the focus of this study and comprise 129 electronic components manufacturers, 70 communication and broadcasting equipment manufacturers, 61 semiconductor manufacturers, 18 video and audio equipment manufacturers, 11 computer and peripheral manufacturers, and 1 magnetic and optical medium manufacturer [41]. Of these listed manufacturers, 126 companies were selected for analysis, based on the publication of their financial reports between 2005 and 2019. These comprise 55 electronic components manufacturers, 32 communication and broadcasting equipment manufacturers, 26 semiconductor manufacturers, 9 video and audio equipment manufacturers, and 4 computer and peripherals manufacturers. Thirty-two of these firms are listed on the KOSPI (Korean Composite Stock Price Index) market, which is the oldest and most representative stock market in Korea. Ninety-four of the firms are listed on the KOSDAQ (Korea Securities Dealers Automated Quotation) market, where the stocks of venture businesses and small and medium-sized enterprises are traded [41]. Considering the trend of firms cutting down on CAPEX during and after the 2019 pandemic [42] but expecting CAPEX to bounce back after a period following any shocks [43], the time span of the study was selected as 2005–2019. This is to exclude the effects of four major events involving a high degree of uncertainty: The 2019–2020 pandemic, the Russian invasion of Ukraine (which started in February 2022), the war in Gaza (which started in October 2023), and the prospect of a global trade war initiated by the Trump administration in 2025.

To test Hypothesis 1, financial results were collected for the period from 2005 to 2019, resulting in balanced panel data comprising 1890 firm-year observations. The data were obtained from filings on the DART (Data Analysis, Retrieval and Transfer System) platform, which is operated by the Financial Supervisory Service in Korea [44]. To test Hypothesis 2, the panel data were divided into two subsets: (A) ‘small and medium-sized enterprises (SMEs)’ and (B) ‘large firms’, depending on firm size. This classification follows the standard classification of the Ministry of SMEs and Startups in Korea [45], which states that enterprises in the electronic components, computers, video and audio equipment, and communication equipment manufacturing industries whose average revenue over the previous three years is less than KRW 100 billion and whose total assets are below KRW 500 billion are classified as small and medium-sized enterprises, while those whose average revenue is above KRW 100 billion or whose total assets are above KRW 500 billion are classified as high-potential and large enterprises.

The summary of the distribution of the firm-year observations is provided in

Table 1. Sample and firm-year distribution by industry subclassification.

Industry Subclassification *	Number of Sample		Total Number in Sample	Number of Observations		Total Number of Observations	Ratio
	SME (Subset A)	Large Firm (Subset B)		SME (Subset A)	Large Firm (Subset B)
Electronic components	28	27	55	420	405	825	44%
Communication and broadcasting equipment	16	16	32	240	240	480	25%
Semiconductor	9	17	26	135	255	390	21%
Video and audio equipment	5	4	9	75	60	135	7%
Computers and peripherals	2	2	4	30	30	60	3%
Total	60	66	126	900	990	1890	100%

Source: calculated by authors. Note: * Subclassification of electronics industry [8].

.

3.2. Methodology

This study uses OLS (Ordinary Least Squares) panel data regression with time-fixed effects to prove the influence of capital expenditure on the financial performance of cross-sectional samples over the years. This is because the fixed-effects regression model removes the influence of omitted time-related variables from the analysis, otherwise it can lead to endogeneity and inconsistent regression model coefficients [46]. Furthermore, many scholars have considered and included fixed effects in OLS regressions to control for time-, firm- or industry-specific effects when examining the impact of capital expenditure on a firm’s profitability [2,6,12,18].

Table 2. Variables and measurement.

Variable		Measurement
Dependent variable	EBITDA ROA	EBITDA divided by total assets
Independent variable	Capital Expenditure (CAPEX)	CAPEX divided by total assets
Control variables	Cash Flow from operations (OCF)	Ratio of cash flow from operations divided by total assets
	Total Asset Growth (TAG)	Total assets change over one year divided by total asset in the previous year
	Asset Turnover (ATO)	Revenue divided by total assets
	Leverage (Lev)	Total debt divided by total assets
	Firm Size (Size)	Dummy variable of firm size; 1 for large firm, 0 for SME

Source: developed by authors.

shows which variables are used in regression and how they are measured. Many previous studies have used NI ROA (net income return on assets) and NI ROE (net income return on equity) as the dependent variable to measure a firm’s profitability [9,10,12,13,31,47,48]. However, this study selects EBITDA ROA (Earnings Before Interest, Taxes, Depreciation, and Amortisation Return on Assets) as a proxy for profitability since net income can be diluted by corporate taxes, interest income/expenses, and depreciation, whereas EBITDA facilitates comparison between firms of different sizes [49]. The independent variable, CAPEX, is measured by the amount of capital expenditure a firm spends in one financial year, divided by its total assets at the end of the year. This approach is based on that of Curtis et al. [18] and Can et al. [50].

Furthermore, five control variables known to influence a firm’s profitability are chosen and added to the regression model, as detailed in the empirical studies: cash flow from operations, total asset growth, asset turnover, leverage, and firm size [4,48,51,52,53,54]. Firstly, cash flow from operations is measured by cash flow from operating activities for one financial year divided by year-end total assets, using the approach of Kim et al. [17] and Can et al. [50]. Secondly, total asset growth is defined as the total asset growth rate over one year [55]. Asset turnover is also calculated, with a high value indicating that the firm is efficiently using assets to generate revenue [15]. Leverage is computed as total debt divided by total assets [17,50,56,57]. Regarding firm size, a value of 1 is assigned to a large firm and a value of 0 is assigned to an SME.

To verify Hypothesis 1, the following panel regression model with a time-fixed effect is established:

EBITDA ROAi,t = β1CAPEXi,t-2 + β2OCFi,t + β3TAGi,t + β4ATOi,t + β5Levi,t + β6Sizei,t +Yeart +εi,t

(1)

where ‘i’ is the individual firm, ‘t’ is the year, ‘Year’ is the time-fixed effect, and ‘ε’ is the error term. This study let a two-year gap between EBITDA ROA and CAPEX and employs the two-year lagged independent variable CAPEXi,t-2 to measure the long-term impact of CAPEX on EBITDA ROA. As can be seen from the previous research, there exists no concrete standard for defining the time length of long-term profitability. Moser et al. [58] defined a time span from one year to five years as long-term, while a period between three years and six years is used as long-term in the research of Turner and Hesford [6]. On the other hand, a one-year and a three-year time lag were used in the study by Kwistianus and Juniarti [9], respectively, to examine the impact of CAPEX on long-term profitability. Therefore, in accordance with IAS 1, which states that assets realised over a period of more than one year are classified as long-term [59] (IFRS Foundation, 2022), a time lag of two years is applied between CAPEX and EBITDA ROA in this study to prove the influence of CAPEX on long-term profitability. In summary, this model measures the effect of a firm’s average CAPEX spending during period t-2 on EBITDA ROA in period t.

After running the regression for Hypothesis 1, this research eliminated the control variable ‘Size’ and implemented additional panel regressions, separately for subsets A (SMEs) and B (large firms), to compare the long-term effects of CAPEX on the ROA of SMEs and large firms, and to prove Hypothesis 2. The regression model used to examine Hypothesis 2 is as follows:

EBITDA ROAi,t = β1CAPEXi,t-2 + β2OCFi,t + β3TAGi,t + β4ATOi,t + β5Levi,t + Yeart + εi,t

(2)

4. Results and Discussion

4.1. Long-Term Impact of CAPEX on EBITDA ROA

The descriptive statistics of the panel data from 2005 to 2019 are presented in

Table 3. Descriptive statistics of panel data.

	All Firms (n * = 1890)			SME (n * = 900)			Large Firms (n * = 990)
Variable	Median	Mean	SD	Median	Mean	SD	Median	Mean	SD
EBITDA ROA	0.070	0.077	0.104	0.052	0.055	0.097	0.086	0.097	0.107
CAPEX **	0.038	0.068	0.081	0.025	0.053	0.067	0.053	0.083	0.090
OCF	0.058	0.062	0.096	0.046	0.047	0.099	0.072	0.076	0.092
TAG	0.051	0.102	0.258	0.040	0.080	0.242	0.059	0.121	0.271
ATO	0.868	0.983	0.608	0.692	0.816	0.561	1.037	1.135	0.609
Lev	0.399	0.394	0.205	0.329	0.346	0.207	0.456	0.438	0.192

Source: Calculated by authors. Note: * number of observations ** 2 year lagged CAPEX.

. Overall, the mean and median values of each variable appear to be higher for large firms than for SMEs. The difference is particularly noticeable in the mean EBITDA ROA value, suggesting that large firms are 1.75 times more profitable than SMEs. Also, as shown in Appendix A, the descriptive statistics results are compared based on the subclassification of the electronics industry. The main finding is that the semiconductor industry demonstrates the highest mean and median EBITDA ROA values and the lowest mean and median asset turnover (ATO) values, implying relatively high profitability compared to other electronics industry sub-sectors, including electronic components, communication and broadcasting equipment, video and audio equipment, and computer and peripheral manufacturing.

Moreover, Figure 1 visualises the cross-sectional mean values of EBITDA, ROA, and two-year lagged CAPEX from 2005 to 2019, which are the focus of this study. This shows that the mean values of EBITDA ROA and CAPEX do not always move in the same direction over time. While they moved in the same direction in 2008, 2011, 2012, 2013, 2014, 2016, 2017, and 2018, they moved in opposite directions in 2006, 2007, 2009, 2010, 2015, and 2019. This suggests that further statistical investigation is needed to verify the correlation between these two variables.

Table 4. Correlation matrix among the independent variable and control variables.

	CAPEXi,t-2	Levi,t	OCFi,t	TAGi,t	ATOi,t	Sizei,t
CAPEXi,t-2	1.000
OCFi,t	0.265	1.000
TAGi,t	0.576	0.212	1.000
ATOi,t	0.239	0.075	0.102	1.000
Levi,t	0.231	0.032	0.172	0.022	1.000
Sizei,t	−0.037	0.186	−0.142	0.062	0.182	1.000

Source: calculated by authors.

shows the correlations between the independent and control variables used in this study. Generally, it is assumed that there is a weak correlation between the variables, except for the correlation between CAPEX and TAG (total asset growth rate). This can lead to multicollinearity, which makes the regression coefficients unreliable [60]. Therefore, the Variance Inflation Factor (VIF) was calculated to detect the multicollinearity among the independent and control variables [61,62].

Panel unit root tests were performed to examine the stationarity of each variable before conducting the regression analysis. These included the Phillips–Perron and Levin–Lin–Chu unit root tests, as presented in Appendix B. According to the results of the Phillips–Perron unit root test, the null hypothesis that the variable is not stationary was rejected, since the absolute values of the test statistics for each variable were greater than the absolute values of the critical values at the 5% significance level and the p-value was less than 0.05 [57]. Based on the results of the Levin–Lin–Chu unit root test, the null hypothesis of a unit root in the variable can be rejected and it can be concluded that all the variables used in the panel regression are stationary. This enables reliable inferences to be made from the regression [25].

In addition, diagnostic tests were conducted to ensure that the assumptions of the fixed-effects regression model were not violated in the time-fixed effects model of this study. The tests checked for the following: (1) no multicollinearity, (2) no heteroscedasticity, (3) no serial correlation, and (4) no cross-sectional dependence. The results of these tests can be found in Appendix C.1, Appendix C.2, Appendix C.3 and Appendix C.4. Firstly, Variance Inflation Factors (VIFs) were calculated to detect multicollinearity among the variables. As the VIF values in Appendix C.1 are all less than 5, the assumption of no multicollinearity is met [60]. Secondly, White’s test was conducted and the residuals were plotted in Appendix C.2 to test for heteroscedasticity. According to the p-value of 0.596, which is greater than 0.05, the null hypothesis of residual homoscedasticity is not rejected [63]. This is evident from the residual dispersion. Thirdly, the Breusch–Godfrey/Wooldridge test was performed to examine serial correlation in the residuals, as shown in Appendix C.3 The p-value (0.000) being lower than 0.05 indicates that the null hypothesis of no serial correlation is rejected, meaning that serial correlation exists in the residuals [23]. Lastly, cross-sectional dependence was diagnosed using the Pesaran CD test (see Appendix C.4). The null hypothesis of no cross-sectional dependence is not rejected because the p-value (0.407) is higher than 0.05. In summary, the time-fixed effects model in this study complies with the assumptions of the absence of multicollinearity, heteroscedasticity, and cross-sectional dependence, except for the assumption of the absence of serial correlation. Therefore, this study uses heteroscedasticity–autocorrelation consistent (HAC) standard errors in the regression to address this violation [13].

To test Hypothesis 1, which states that capital expenditure positively influences long-term profitability, regression analyses were performed using time-fixed effects, random-effects, and pooled regression models. This was carried out to ensure consistency in the coefficients of the independent and control variables across the different types of panel regression model. The regression results from the time-fixed effects model, the random-effects model, and the pooled regression model are reported in

Table 5. Hypothesis 1 testing: Regression results from the fixed-effects model, random-effects model, and pooled regression model.

Independent/Control Variables	Dependent Variable: EBITDA ROAi,t
	Time-Fixed Effects	Random Effects	Pooled Regression
CAPEXi,t-2	0.171 *** (7.062)	0.155 *** (6.961)	0.180 *** (8.148)
OCFi,t	0.529 *** (17.524)	0.490 *** (15.976)	0.532 *** (17.513)
TAGi,t	0.066 *** (6.670)	0.070 *** (7.882)	0.070 *** (7.697)
ATOi,t	0.021 *** (4.835)	0.029 *** (6.007)	0.022 *** (5.382)
Levi,t	−0.023 ** (−2.495)	−0.029 *** (−2.750)	−0.021 * (−2.340)
Sizei,t	0.014 *** (3.427)	0.013 ** (2.475)	0.013 ** (3.124)
Intercept	0.007 (1.307)	0.005 (0.874)	0.004 (0.861)
R2	0.390	0.363	0.406
Adjusted R2	0.383	0.361	0.404
F-statistic	198.750 *** (d.f. = 6; 1883)	-	214.222 *** (d.f. = 6; 1883)
Chi-square		1073.99 *** (d.f. = 6)
Observation	1890	1890	1890

Source: Author’s own work Note: t-value in parentheses. * significant at 10% level. ** significant at 5% level. *** significant at 1% level.

.

First of all, it is assumed that the dependent variable EBITDA ROAi,t is moderately explained by the independent variable CAPEXi,t-2. Secondly, the independent variable CAPEXi,t-2 is shown to be positively correlated with the dependent variable EBITDA ROAi,t, as consistently witnessed from the coefficients of CAPEXi,t-2 under the fixed-effects model (0.171), random-effects model (0.155) and pooled regression model (0.180). Moreover, these positive coefficients of CAPEXi,t-2 are significant at the 1% significance level, which is evidence to support Hypothesis 1. From a managerial perspective, this suggests that capital expenditure takes time to have a financial impact, so firms in capital-intensive industries should consider CAPEX decisions in terms of medium-term profitability rather than immediate outcomes.

Also, each control variable appears to have significant correlation with the dependent variable EBITDA ROAi,t. OCFi,t, TAGi,t, ATOi,t, and Sizei,t are shown to be positively related to EBITDA ROAi,t, whereas Levi,t has a negative relationship with EBITDA ROAi,t. Looking at the results of time-fixed effects model, first of all, it is demonstrated that the coefficient of Sizei,t is 0.014, implying the improved profitability due to economies of scale [46,51] and supporting the results of the previous study by Kwistianus and Juniarti [9], Rahman and Yilun [64], Etim [48] and Singh et al. [12]. The positive correlation between OCFi,t (cash flow from operations) and EBITDA ROAi,t, consistent with the outcome from the study by Kim et al. [17], is 0.529, which shows the most contribution among the variables, and denotes that 1% increase in cash flow from operations leads to a 0.529% increase in EBITDA ROAi,t. Also, TAGi,t (total asset growth rate) positively affects EBITDA ROAi,t, (0.066), as shown in the research by Inyiama et al. [53]. In addition, ATOi,t (asset turnover) shows a positive relationship (0.021) with EBITDA ROAi,t as well, which is aligned with the previous literature in that an increase in asset turnover contributes to enhancing profitability by efficiently making use of the assets [23]. However, Levi,t is found to be negatively associated (−0.023) with EBITDA ROAi,t, corresponding with the study of Kwistianus and Juniarti [9], Yadav et al. [31], which suggests that the samples of this study do not show the improvement of the agency cost while they increase the amount of debt [56]. The practical implication is that firms should carefully balance their investment financing, since excessive leverage can undermine the positive impact of CAPEX on long-term profitability. Overall, while the empirical results remain central to the analysis, their interpretation highlights important implications for economic and financial management. The findings indicate that long-term profitability depends not only on CAPEX itself but also on supporting conditions such as liquidity strength, efficient asset use, sustainable financing structures, and the ability to realise economies of scale.

In summary, the regression models in this study verify the positive correlation between two-year lagged CAPEX (CAPEX_(it), t-2) and EBITDA ROA, thus confirming Hypothesis 1, which states that CAPEX has a positive influence on a firm’s long-term profitability. An F-test and a Hausman test were conducted to ensure that the time-fixed effects model was more appropriate than the random-effects model and the pooled regression model for estimating the correlation between CAPEX and EBITDA ROA. Firstly, an F-test was carried out to compare the time-fixed effects model and the pooled regression model [46]. According to the results in Appendix D.1, it can be concluded that a significant time effect exists, meaning that the time-fixed effects model is more appropriate than the pooled regression model for controlling the time effect. Secondly, a Hausman test was conducted to examine endogeneity in the random-effects model and to choose between the random-effects model and the time-fixed effects model [46]. The Hausman test in Appendix D.2 confirms that the random-effects model is inconsistent and that the time-fixed effects model is more appropriate. In summary, the F-test and the Hausman test show that the time-fixed effects model is more appropriate than the random-effects model and the pooled regression model for this study. This lends credence to the decision to apply the fixed-effects model to analyse the panel data.

Furthermore, the robustness of the regression results is tested to ensure the validity of the coefficients for the independent and control variables in the time-fixed effects model [65]. Firstly, the robustness of the time-fixed effects model is examined by changing the time lag between independent variable CAPEX and the dependent variable EBITDA ROA. In detail, the regression was carried out using one-year and three-year lagged CAPEX.

Table 6. Robustness test 1 of Model 1.

	Dependent Variable: EBITDA ROAi,t
	(1) 1-Year Lag	(2) 2-Year Lag	(3) 3-Year Lag
CAPEXi	0.159 *** (6.814)	0.171 *** (6.784)	0.150 *** (6.876)
OCFi,t	0.533 *** (17.948)	0.529 *** (25.558)	0.532 *** (17.797)
TAGi,t	0.064 *** (6.510)	0.066 *** (8.841)	0.066 *** (6.520)
ATOi,t	0.020 *** (4.761)	0.021 *** (6.225)	0.021 *** (4.861)
Levi,t	−0.022 ** (−2.231)	−0.023 * (−2.312)	−0.022 ** (−2.245)
Sizei,t	0.014 *** (3.398)	0.014 *** (3.499)	0.015 *** (3.518)
Intercept	0.008 (1.580)	0.007 (1.374)	0.007 (1.355)
R2	0.387	0.390	0.388
Adjusted R2	0.385	0.383	0.386
F-statistic	196.365 *** (d.f. = 6; 1883)	198.750 *** (d.f. = 6; 1883)	197.067 *** (d.f. = 6; 1883)
Observation	1890	1890	1890

Source: calculated by authors. Note: t-value in parentheses. CAPEXi,t-1, CAPEXi,t-2, and CAPEXi,t-3 are used as the independent variable for the model’s (1) 1-year lag, (2) 2-year lag, and (3) 3-year lag, respectively. * significant at 10% level. ** significant at 5% level. *** significant at 1% level.

shows the regression results for one-year and three-year lagged CAPEX, as well as two-year lagged CAPEX, which was originally used in the first regression model in this study. The overall coefficients for the independent variable, CAPEX, and the control variables are consistent across models with different time lags for CAPEX. Also, the significance of the coefficients remains unchanged, and the R² value from the model with a two-year lagged CAPEX (0.390) is almost identical to the R² values from the models with a one-year (0.387) and three-year (0.388) lagged CAPEX. Therefore, the regression results from the time-fixed effects model in Table 5 are considered robust. Additionally, this robustness test provides evidence that CAPEX has a positive impact on profitability in the short and long term, implying that the influence of CAPEX lasts over the years.

The second robustness test involves removing all the control variables from the regression model and then adding them back in one by one [66], while observing the changes in the value and significance of the coefficients and R² of each model. As shown in

Table 7. Robustness test 2 of Model 1.

	Dependent Variable: EBITDA ROAi,t
	(1)	(2)	(3)	(4)	(5)	(6)
CAPEXi,t-2	0.311 *** (9.734)	0.169 *** (6.749)	0.166 *** (6.789)	0.175 *** (7.065)	0.184 *** (7.577)	0.171 *** (7.062)
OCFi,t		0.583 *** (19.298)	0.569 *** (19.184)	0.545 *** (18.021)	0.537 *** (17.856)	0.529 *** (17.524)
TAGi,t			0.066 *** (6.516)	0.067 *** (6.716)	0.068 *** (6.850)	0.066 *** (6.670)
ATOi,t				0.022 *** (5.268)	0.023 *** (5.422)	0.021 *** (4.835)
Levi,t					−0.016 * (−1.782)	−0.023 ** (−2.495)
Sizei,t						0.014 *** (3.427)
Intercept	0.056 *** (16.606)	0.029 *** (9.791)	0.024 *** (7.764)	0.003 (0.551)	0.008 (1.424)	0.007 (1.307)
R2	0.056	0.343	0.369	0.385	0.385	0.390
Adjusted R2	0.055	0.342	0.368	0.383	0.384	0.383
F-statistic	110.404 *** (d.f. = 1; 1888)	488.715 *** (d.f. = 2; 1887)	364.106 *** (d.f. = 3; 1886)	292.289 *** (d.f. = 4; 1885)	234.64 *** (d.f. = 5; 1884)	198.750 *** (d.f. = 6; 1883)
Observation	1.890	1.890	1.890	1.890	1.890	1.890

Source: calculated by authors. Note: t-value in parentheses. * significant at 10% level. ** significant at 5% level. *** significant at 1% level.

, the value of the CAPEX_(it)-2 coefficient and the R² value of the model change considerably when OCF (cash flow from operations) is added as a control variable: the CAPEX_(it)-2 coefficient moves from 0.311 to 0.169, and the R² value of the regression model changes from 0.056 to 0.343. This indicates that the control variable OCF (cash flow from operations) has a considerable effect on the dependent variable EBITDA ROA. However, the coefficients of the variables and the R² value of the model remain consistent afterwards. In summary, the regression results from the time-fixed effects model prove Hypothesis 1 and are assumed to be robust given the consistent values and significance of the variable coefficients.

4.2. Comparison of Long-Term Impact of CAPEX on EBITDA ROA Between SME and Large Firm

To test Hypothesis 2, the panel data is divided into two subsets: subset A comprising SMEs with 900 observations and subset B comprising large firms with 990 observations. Regression analysis is conducted separately on subsets A and B, and the CAPEX coefficients from the two subsets are compared. The study follows the same analysis procedure as that applied to verify Hypothesis 1: first, regression is carried out using a time-fixed effects model, a random-effects model, and a pooled regression model. Secondly, an F-test and a Hausman test are conducted to confirm that the fixed-effects model is more appropriate than the random-effects model and the pooled regression model for each subset. Thirdly, a T-test is performed to determine whether the impact of CAPEX on large firms (subset B) is significantly greater than the impact on SMEs (subset A). Additionally, robustness tests are conducted to examine the robustness of the results.

Firstly, panel unit root tests (the Phillips–Perron and Levin–Lin–Chu unit root tests) were conducted separately for two subsets to check the stationarity of the variables. As shown in Appendix E, it was concluded that there was no unit root in either subset for each variable. Diagnostic tests were also implemented to determine whether the assumptions of the time-fixed effects model were met, as demonstrated in Appendix F.1, Appendix F.2, Appendix F.3, Appendix F.4, Appendix G.1, Appendix G.2, Appendix G.3 and Appendix G.4. According to the test results (see Appendix F.1, Appendix F.2, Appendix F.3 and Appendix F.4), the time-fixed effects model for SMEs complies with the assumptions of the absence of multicollinearity, heteroscedasticity, and cross-sectional dependence, but not with the assumption of the absence of serial correlation. In the time-fixed effects model of the large firm (see Appendix G.1, Appendix G.2, Appendix G.3 and Appendix G.4), multicollinearity and heteroscedasticity do not exist, whereas serial correlation and cross-sectional dependence are present in the residuals. In summary, the diagnostic tests show that the time-fixed effects model of SME does not comply with the assumption of no serial correlation, and that the model of the large firm does not comply with the assumptions of no serial correlation or cross-sectional dependence. Therefore, heteroscedasticity-autocorrelation consistent (HAC) standard errors will be used in the regression to overcome the presence of serial correlation and cross-sectional dependence [67].

Table 8. Hypothesis 2 testing: Regression results from the fixed-effects model, random-effects model, and pooled regression model.

	Dependent Variable: EBITDA ROAi,t
	Time-Fixed Effects		Random Effects		Pooled Regression
	SME	Large Firm	SME	Large Firm	SME	Large Firm
CAPEXi,t-2	0.150 *** (3.925)	0.187 *** (5.348)	0.137 *** (3.716)	0.168 *** (5.952)	0.166 *** (4.471)	0.191 *** (6.884)
OCFi,t	0.511 *** (12.056)	0.535 *** (16.404)	0.478 *** (10.986)	0.492 *** (11.384)	0.520 *** (11.963)	0.536 *** (12.819)
TAGi,t	0.084 *** (6.933)	0.054 *** (4.980)	0.090 *** (7.218)	0.055 *** (4.887)	0.088 *** (7.094)	0.058 *** (4.852)
ATOi,t	0.018 *** (3.022)	0.023 *** (4.748)	0.029 *** (3.752)	0.029 *** (4.850)	0.021 *** (3.323)	0.024 *** (4.367)
Levi,t	−0.020 (−1.614)	−0.026 * (−1.716)	−0.029 ** (−2.009)	−0.029 * (−1.797)	−0.016 (−1.310)	−0.026 * (−1.935)
Intercept	0.008 (1.415)	0.020 ** (2.225)	0.004 (0.6147)	0.018 * (1.962)	0.003 (0.607)	0.017 ** (2.081)
R2	0.419	0.322	0.428	0.291	0.449	0.334
Adjusted R2	0.416	0.319	0.425	0.288	0.446	0.330
F-statistic	127.030 *** (d.f. = 5; 894)	92.295 *** (d.f. = 5; 984)	-	-	145.54 *** (d.f. = 5; 894)	98.457 *** (d.f. = 5; 984)
Chi-square			670.009 *** (d.f. = 5)	404.357 *** (d.f. = 5)
Observation	900	990	900	990	900	990

Source: Author’s own work. Note: t-value in parentheses. * significant at 10% level. ** significant at 5% level. *** significant at 1% level.

shows the regression results for the impact of CAPEX on a firm’s long-term profitability using three models: the fixed-effects model, the random-effects model and the pooled regression model. Firstly, the R² values from the different regression models show that EBITDA ROAi,t is adequately explained by CAPEXi,t-2 [68]. Secondly, the positive influence of CAPEXi,t-2 on EBITDA ROAi,t is demonstrated for both SMEs and large firms, as shown by the time-fixed effects model: 0.187; random-effects model: 0.168; pooled regression model: 0.191; time-fixed effects model: 0.187; random-effects model: 0.168; pooled regression model: 0.150. The influence of SMEs is weaker than that of large firms: 0.150; random-effects model: 0.137; pooled regression model: 0.166, as evidenced by the significant coefficients of CAPEX_(it)-2 at the 1% significance level. This provides empirical support for the argument that larger firms benefit more from capital investments due to their scale advantages and greater ability to absorb long-term projects. Additionally, the overall control variables have a significant effect on EBITDA ROAi,t, which is consistent with the results of the regression with the entire panel data. The effect of large firms is stronger than that of SMEs, except for TAG (total asset growth rate). In summary, the regression models used to analyse the SME and large firm subsets imply that two-year lagged CAPEX (CAPEX_(it)-2) is positively associated with EBITDA ROA_(it) for both SMEs and large firms. Also, the coefficient of the two-year lagged CAPEX (CAPEX_(i,t-2)) for large firms is larger than that for SMEs, and further analysis is required to verify whether this difference is significant.

From a practical economic management perspective, these results highlight the challenges that firms of different sizes face when implementing capital investment strategies. While large firms benefit from economies of scale and stronger financial capacity, they also face coordination complexity, longer decision-making cycles and the risk of bureaucratic inefficiencies. SMEs, on the other hand, often encounter financing constraints, higher capital costs, and limited cash reserves, which can restrict their ability to sustain multi-year investment programmes. The weaker effect of CAPEX among SMEs may therefore reflect structural limitations in funding, risk-bearing capacity, and the efficiency with which assets are utilised.

Furthermore, SMEs’ stronger dependence on internal cash flow and asset growth highlights a common managerial issue in smaller enterprises: balancing short-term liquidity pressures with long-term investment objectives. In contrast, large firms may struggle to ensure that capital investments translate into genuine efficiency gains rather than reinforcing existing inefficiencies. Thus, while the statistical results show positive associations, they also reveal practical challenges in economic management, such as maintaining investment discipline, financial flexibility, and aligning CAPEX decisions with operational capabilities.

According to the results of the F-test and the Hausman test in Appendix H.1 and Appendix H.2, it is suggested that the time-fixed effects model is more appropriate for estimating the correlation between EBITDA, ROAi and the explanatory variables, including CAPEXi, for SMEs. For large firms, the pooled regression model is sufficient for estimating the coefficients, as shown in Appendix I.1 and Appendix I.2. Nevertheless, to facilitate comparison of the CAPEX_(it-2) coefficients between SMEs and large firms, this study applies the time-fixed effects model to the large firm, using the same regression methodology as for the SMEs. In other words, this study compares the CAPEX_(it-2) coefficient of SMEs with the CAPEX_(it-2) coefficient of large firms using a time-fixed effects model to prove Hypothesis 2, which states that large firms show a stronger impact of capital expenditure on long-term profitability than small and medium-sized firms.

Table 9. Testing Hypothesis 2: Coefficients difference verification with 2-year lag (T-test).

Independent Variable	Coefficient Difference	Standard_Error Difference	t_Value	p_Value
CAPEXi,t-2	0.037	0.052	0.707	0.519

Source: calculated by authors. Note: Null hypothesis: The difference in the coefficients of CAPEXi,t-2 between SMEs and large firms is zero.

shows the results of the T-test conducted to prove the significance of the difference in the CAPEXi,t-2 coefficients between SMEs and large firms. According to the p-value (0.519), which is higher than 0.05, the null hypothesis that the CAPEXi,t-2 coefficient for SMEs is not different from that for large firms is not rejected. Therefore, it can be concluded that the difference in the impact of CAPEXi,t-2 on EBITDA ROAi,t between SMEs and large firms is not significant, providing insufficient evidence to support Hypothesis 2.

In addition, a robustness test is implemented to verify the reliability of the regression results. Firstly, the robustness is examined by changing the time lag between the predictor variable CAPEX and the dependent variable EBITDA ROA from one year to three years using a time-fixed effects model. Overall, the coefficients of the predictor variable CAPEX, as well as those of the control variables, are consistent across the different time-lagged regression models, as shown in

Table 10. Robustness test 1 of Model 2.

	Dependent Variable: EBITDA ROAi,t
	SME	Large Firm	SME	Large Firm	SME	Large Firm
	(1) 1-Year Lag		(2) 2-Year Lag		(3) 3-Year Lag
CAPEXi	0.123 *** (3.870)	0.188 *** (6.008)	0.150 *** (3.925)	0.187 *** (5.348)	0.137 *** (4.402)	0.157 *** (5.134)
OCFi,t	0.512 *** (11.954)	0.541 *** (13.007)	0.511 *** (12.056)	0.535 *** (16.404)	0.512 *** (11.879)	0.543 *** (12.928)
TAGi,t	0.083 *** (6.727)	0.050 *** (3.753)	0.084 *** (6.933)	0.054 *** (4.980)	0.082 *** (6.635)	0.055 *** (3.868)
ATOi,t	0.018 *** (2.878)	0.022 *** (4.004)	0.018 *** (3.022)	0.023 *** (4.748)	0.019 *** (3.003)	0.022 *** (3.940)
Levi,t	−0.019 ** (−1.496)	−0.026 * (−1.860)	−0.020 (−1.614)	−0.026 * (−1.716)	−0.021 * (−1.660)	−0.023 (−1.587)
Intercept	0.010 * (1.714)	0.021 ** (2.233)	0.150 *** (3.925)	0.187 *** (5.348)	0.009 (1.452)	0.020 ** (2.199)
R2	0.415	0.322	0.419	0.322	0.419	0.318
Adjusted R2	0.412	0.318	0.416	0.319	0.416	0.314
F-statistic	125.032 *** (d.f. = 5; 894)	92.017 *** (d.f. = 5; 984)	127.030 *** (d.f. = 5; 894)	92.295 *** (d.f. = 5; 984)	127.046 *** (d.f. = 5; 894)	90.416 *** (d.f. = 8; 984)
Observation	900	990	900	990	900	990

Source: Author’s own work. Note: t-value in parentheses. CAPEXi,t-1, CAPEXi,t-2, and CAPEXi,t-3 are used as independent variables for the model’s (1) 1-year lag, (2) 2-year lag, and (3) 3-year lag, respectively. * significant at 10% level. ** significant at 5% level. *** significant at 1% level.

. Also, the CAPEX coefficients for large firms with a one-year time lag (0.188), two-year time lag (0.187) and three-year time lag (0.157) remain steadily larger than those for SMEs (0.123 with a one-year time lag, 0.150 with a two-year time lag, and 0.137 with a three-year time lag). Furthermore, the significance of the coefficients remains significant, and the R² values of the models with different time lags remain constant. Therefore, it can be assumed that the regression results of the time-fixed effects model for SMEs and large firms in Table 8 are robust. Additionally, a T-test was performed to examine whether the CAPEXi,t-1 and CAPEXi,t-3 coefficients of SMEs differ significantly from those of large firms.

Table 11. Verification of the difference in coefficients with 1-year lag and 3-year lag (T-test).

Independent Variable	Coefficient Difference	Standard_Error Difference	t_Value	p_Value
CAPEXi,t-1	0.065	0.054	1.210	0.293
CAPEXi,t-3	0.020	0.048	0.410	0.703

Source: calculated by authors. Note: Null hypothesis: The differences in the coefficients of CAPEXi,t-1 and CAPEXi,t-3 between SMEs and large firms are zero.

shows the results of the t-test indicating that the difference between the CAPEXi,t-1 and CAPEXi,t-3 coefficients for SMEs and large firms is not significant. This suggests that there is no significant difference in the impact of CAPEX on short-term and longer-term profitability between SMEs and large firms.

The second robustness test involves eliminating the control variables from the original regression model with the independent variable ‘2-year lagged CAPEX’ and then adding each control variable back to the model.

Table 12. Robustness test 2 of Model 2 for SME and large firms.

	Dependent Variable: EBITDA ROAi,t
	(1)	(2)	(3)	(4)	(5)
	SME
CAPEXi,t-2	0.205 *** (3.949)	0.130 *** (3.194)	0.141 *** (3.582)	0.143 *** (3.680)	0.150 *** (3.925)
OCFi,t		0.569 *** (13.169)	0.536 *** (13.100)	0.522 *** (12.844)	0.511 *** (12.056)
TAGi,t			0.081 *** (6.516)	0.084 *** (6.919)	0.084 *** (6.933)
ATOi,t				0.017 *** (2.980)	0.018 *** (3.022)
Levi,t					−0.020 (−1.614)
Intercept	0.044 *** (9.536)	0.022 *** (5.613)	0.016 *** (4.209)	0.002 (0.431)	0.150 *** (3.925)
R2	0.020	0.367	0.408	0.418	0.419
Adjusted R2	0.018	0.366	0.406	0.415	0.416
F-statistic	17.538 *** (d.f. = 1;898)	256.121 *** (d.f. = 2;897)	202.326 *** (d.f. = 3;896)	157.846 *** (d.f. = 4;895)	127.030 *** (d.f. = 5; 894)
Observation	900	900	900	900	900
Large firms
CAPEXi,t-2	0.307 *** (7.867)	0.155 *** (5.041)	0.149 *** (5.003)	0.173 *** (5.349)	0.187 *** (5.348)
OCFi,t		0.568 *** (12.786)	0.569 *** (13.042)	0.547 *** (12.121)	0.535 *** (16.404)
TAGi,t			0.051 *** (3.6197)	0.052 *** (3.747)	0.054 *** (4.980)
ATOi,t				0.021 *** (3.561)	0.023 *** (4.748)
Levi,t					−0.026 * (−1.716)
Intercept	0.071 *** (15.150)	0.041 *** (8.535)	0.035 *** (7.088)	0.010 (1.280)	0.187 *** (5.348)
R2	0.062	0.290	0.306	0.320	0.322
Adjusted R2	0.061	0.289	0.304	0.318	0.319
F-statistic	64.416 *** (d.f. = 1; 988)	198.961 *** (d.f. = 2; 987)	142.960 *** (d.f. = 3; 986)	114.403 *** (d.f. = 4; 985)	92.295 *** (d.f. = 5; 984)
Observation	990	990	990	990	990

Source: calculated by authors. Note: t-value in parentheses. * significant at 10% level. ** significant at 5% level. *** significant at 1% level.

shows that, for both SMEs and large firms, the coefficients of the independent and control variables and the significance of the coefficients remain consistent, with no considerable change, after the control variable OCF (cash flow from operations) is added. This variable contributes most to EBITDA ROAit. Thus, the results of the time-fixed effects model for SMEs and large firms are considered robust according to the consistency and significance of the coefficients.

The findings showed that capital expenditure has a positive effect on a firm’s long-term profitability, providing evidence to support Hypothesis 1. This result is consistent with previous research demonstrating that capital expenditure improves long-term profitability [9]. Additionally, this study found that capital expenditure is positively correlated with short-term and long-term profitability, implying that its impact is effective and long-lasting. This outcome is consistent with the findings of Moser et al. [58]. Furthermore, the long-lasting influence of capital expenditure explains how it contributes to a firm’s profitability over time, providing empirical evidence as to why CAPEX is capitalised and depreciated over several years. It was also found that the impact of CAPEX on large firms is greater than on SMEs, but the difference is not statistically significant; therefore, Hypothesis 2 is not supported. Diseconomies of scale could be a plausible reason why the impact of CAPEX on large firms is similar to that on SMEs in this study. As firms expand and produce above their optimum level of output, their profitability starts to decrease due to additional costs [44]. This additional cost is primarily due to the increased number of decision-making levels, which makes it difficult to react quickly to changes in the market situation due to the longer decision-making process [51]. Additionally, the additional costs stem from the agency problem, whereby managers prioritise their own benefits over maximising stakeholder value and firm profits [54].

5. Conclusions

Drawing on 1890 observations from 126 listed electronics firms in South Korea between 2005 and 2019, this study investigated how the impact of capital expenditure on long-term profitability differs depending on firm size. Motivated by inconsistent findings in prior literature regarding the effectiveness of CAPEX and the role of firm size, two hypotheses were formulated: (1) capital expenditure positively affects long-term profitability, and (2) large firms benefit more from capital expenditure than small and medium-sized enterprises (SMEs). To test these hypotheses, a time-fixed effects panel regression model was employed using EBITDA ROA in year t as the dependent variable and CAPEX in year t–2 as the core explanatory variable. Operating cash flow, total asset growth, asset turnover, leverage, and firm size were used as control variables. Additional regressions were then conducted on SME and large-firm subsamples.

The empirical findings confirm that CAPEX has a positive and statistically significant effect on long-term profitability, thus supporting Hypothesis 1.

Although the CAPEX coefficient is higher among large firms than SMEs, the difference is not statistically significant. This leads to the rejection of Hypothesis 2, suggesting that scale does not automatically amplify the impact of capital expenditure on profitability. In fact, the absence of a significantly stronger effect among large firms is consistent with the presence of potential diseconomies of scale, such as coordination inefficiencies, lower marginal returns on investment or more complex organisational structures, that may weaken the effectiveness of additional capital outlays. Furthermore, the multi-year lag structure observed in the results reflects the long implementation and adjustment periods inherent in capital budgeting processes.

5.1. Practical Implications

The findings have several important implications for managers and policymakers in capital-intensive industries. Firstly, CAPEX should be evaluated from a long-term rather than a short-term perspective. The positive two-year lagged effect indicates that investment projects require time to influence profitability; therefore, excessive focus on immediate earnings may hinder necessary strategic investments. Secondly, SMEs should be wary of underinvesting due to liquidity constraints, since the results demonstrate that even smaller firms can achieve significant long-term profitability gains through sustained capital investment. Policies that facilitate access to affordable financing could therefore help SMEs to increase their productivity. Thirdly, the absence of a stronger CAPEX effect in large firms signals the need for disciplined investment governance. Large firms must ensure that organisational inefficiencies do not dilute the impact of capital projects, and must strengthen monitoring, accountability, and asset utilisation practices. Fourthly, the strong positive effects of internal cash flow, asset turnover, and firm size emphasise the importance of financial flexibility and operational efficiency in enabling investment effects to materialise. Finally, to avoid short-termism and ensure that capital investments support sustainable growth, firms should align managerial incentives with long-term value creation.

5.2. Limitations and Directions for Future Research

This study has several limitations that should be addressed in future research. Firstly, the sample only includes listed electronics firms in South Korea, which limits the generalisability of the findings to the broader electronics industry and to unlisted firms that may follow different investment and financing patterns. Secondly, the study covers the relatively stable macroeconomic period of 2005–2019. However, the period after 2019 has been characterised by major structural shocks, including the ongoing geopolitical conflicts, the rapid shifts in US trade and industrial policy during and after the Trump administration, the global supply-chain instability, and the pandemic. These disruptions have fundamentally altered investment cycles, financing conditions, and profitability dynamics. Analysing CAPEX effects in such a context would require different methodological approaches, such as structural break tests, threshold models, or time-varying parameter models, in order to capture nonlinearities and regime shifts. Thirdly, this study does not differentiate between types of fixed assets, despite their useful lives, technology intensity, and returns on investment differing significantly. Future research should classify CAPEX by asset category to identify which types of investments generate the strongest and most persistent profitability effects. Finally, extending the dataset to include longer time spans or incorporating more granular operational data could provide deeper insights into the duration, peak, and decay patterns of CAPEX impacts on firm performance.