The Impact of Selected Financial Ratios on Economic Value Added: Evidence from Croatia

Abstract

Traditional financial performance measures should be extended to provide additional information to stakeholders. One such extension is the economic value added (EVA). It shows residual profit above the cost of financing, both creditors and equity financing. This paper elaborates on the impact of selected financial ratios on EVA to total assets and EVA to capital employed using the 20-year aggregated data of non-financial business entities operating in Croatia. It answers the research question of which of the selected financial ratios impacts the above-mentioned EVA-based ratios. Applying dynamic panel data modeling using the generalized method of moments technique resulted in the derivation of two models. The human capital efficiency ratio was statistically significant in both models, positively affecting EVA/total assets and EVA/capital employed. In contrast, the debt ratio and net profit margin were significant only in the second model, where EVA/capital employed was a dependent variable. The research results indicate that the debt ratio affects EVA/capital employed negatively while the net profit margin has a positive effect, confirming the existing research. Total liabilities/earnings before interest, taxes, depreciation and amortization, and total asset turnover were not found to be significant in either of the two models.

1. Introduction

Traditional financial performance measures include only the costs of lenders’ capital in calculating companies’ overall profitability. They overlook that shareholders are engaging their sources of financing to obtain a rate of return. Economic value added (EVA) represents a contemporary measure of financial performance that shows the economic profit available to owners (Zenzerović 2023). It is a category that provides information on the residual profit above the costs of the sources of financing, and as such represents the extension of classic financial performance measures such as net operating profit after tax (NOPAT), earnings before interest, tax, amortization, and depreciation (EBITDA), and the accounting result, i.e., net income or loss (Zenzerović 2023). This paper aims to determine the impact of selected financial ratios on the EVA using the aggregated data for a 20-year period of non-financial business entities operating in the Republic of Croatia. Instead of using EVA’s absolute value, the ratios EVA/total assets (EVA_TA) and EVA/capital employed (EVA_CE) are employed. According to the authors’ knowledge, there are scarce research results that elaborate on the relationship between financial ratios and selected EVA-based indicators as well as the ones that analyze the whole population of business entities in one economy. This is the reason the authors use the EVA-based indicators that show the relations between the EVA and selected accounting variables. This approach results in calculating the specific efficiency ratios. The problem the authors want to solve consists of estimating the impact of selected financial ratios on the new EVA-based indicators.

EVA represents an analytical tool commercially developed in 1982 by Joel Stern and G. Bennett Stewart as a result of several years of research started in the 1970s (Grant 2003). Alfred Marchall, a famous British economist, was the one who first wrote about the concept of total costs of financing at the beginning of the twentieth century. A couple of years later, the EVA was implemented by General Motors (Zenzerović 2023).

EVA represents a measure of the dollar surplus value created by an investment or a portfolio of investments. It is computed as the product of the excess return made on an investment or investments and the capital invested in that investment or investments (Damodaran 2012).

Equations (1) and (2) show the calculation of EVA (Zenzerović 2023).

EVA = (ROI − WACC) × CI

(1)

EVA = NOPAT − (WACC × CI)

(2)

where:

• EVA = economic value added;
• ROI = return on capital invested;
• WACC = weighted average cost of capital;
• CI = capital invested, represented by the sum of capital and reserves and financial liabilities;
• NOPAT = net operating profit after taxes.

The weighted average cost of capital calculation is shown in Equation (3) (Dobrowolski et al. 2022).

WACC = (E/V × Re) + (D/V × Rd × (1 − Tc)

(3)

where:

• WACC = weighted average cost of capital;
• E = market value of the company’s equity;
• D = market value of the company’s debt;
• Re = cost of equity;
• Rd = cost of debt;
• Tc = corporate tax rate.

The EVA methodology is the one measure that properly accounts for all the complex trade-offs involved in creating value, and therefore, is the right measure to use for setting goals, evaluating performance, determining bonuses, communicating with investors, and for capital budgeting and valuations of all sorts (Stewart 1991). It became popular as an awarding system, system of strategic planning, and value-based management system that directs a company’s organizational behavior. Nowadays, EVA is implemented to determine managerial bonuses. It is sometimes used as a measure of financial performance. The introduction of the EVA index as a performance indicator in state-owned companies in China resulted in an increase in the level of money in companies and a reduction in excessive investment activities (Shen et al. 2015). While the introduction of EVA as a performance measurement tool is generally considered consistent with mitigating agency costs, and therefore, increasing shareholder value, these same actions can also be associated with sub-optimal decisions (e.g., reducing investment in positive NPV projects to avoid the now explicit capital charge) (Wallace 1997). That is why EVA must be examined in a broader context along with other financial performance indicators, because managers may continue the practice of pitting their short-term goals against the long-term goals of the owners (Zenzerović 2023).

The main disadvantage of EVA is that it is still a financial performance measurement tool. These tools rely mostly on accounting earnings that often fail to measure changes in the economic value of the company. Sabol and Sverer found that the main reason for this lies in the fact that the accounting earnings ignore the time value of money, they do not consider dividend policy, they represent the accrual-based values that are different from cash flows from operations, and they do not include business and financial risk. Last but not least is that accounting earnings depend on alternative accounting methods that may be employed, resulting in different performances (Sabol and Sverer 2017). Additional disadvantages of EVA were emphasized by Cinotti, who states that EVA ignores investment in business continuity, is complex to implement, results in problems with the weighted average cost of capital calculation, and focuses on short-term objectives (Cinotti 2023).

EVA’s absolute value usually varies significantly among different activities and periods and can be misleading as a financial ratio. This is why the research used the relation between EVA and total assets (EVA_TA) and EVA to capital employed (EVA_CE). Such ratios give more information on the efficiency of total assets and capital and reserves used. The next step was to determine the financial ratios that impact the EVA-based ratios. In this research, the solvency, activity, profitability, and human capital efficiency ratios were employed. Solvency ratios included total liabilities to EBITDA (TL_EBITDA) and debt ratio (DR), the total assets turnover (TAT) represented activity ratios, while for a profitability measurement the net profit margin (NPM) was used. The authors selected these ratios according to their frequent application in estimating the financial characteristics of companies. Existing research indicates that the relation between the solvency ratio and EVA is negative because the increased level of debt leads to higher interest expenses, reducing the residual value available to shareholders and decreasing the EVA (Winarko and Jaya 2018; Utam and Purnamasari 2023; Novyarni and Nur Ayu Ningsih 2019). The TAT ratio indicates the efficiency of asset utilization, which can positively impact EVA as well as NPM, which indicates the profitability that is positively related to EVA (Duhita and Rizkianto 2023; Utam and Purnamasari 2023; Novyarni and Nur Ayu Ningsih 2019). Intellectual capital has become a keyword in the knowledge-based economy, so we decided to test the impact of the human capital efficiency ratio on EVA-based ratios. Research results indicate that higher human capital efficiency can positively influence EVA by increasing the operating profits and residual value available to shareholders (Dash et al. 2013; Okoye and Emeneka 2021; Mohammadi Cheshmeh Kaboudi et al. 2015).

The research conducted aimed to answer the following research question:

Which of the selected financial ratios impact the EVA_TA and EVA_CE?

2. Methodology and Data

The primary goal of this research is to assess the impact of selected financial ratios on the EVA-based ratios of Croatian companies. To accomplish this, two dynamic panel data models are estimated using the generalized method of moments (GMM) technique. Dynamic panel models are better than static panel models for several reasons. For example, they are more effectively specified and incorporate dynamics in the estimated portion of the model, rather than attributing them to the error term which would compromise fixed- or random-effect estimation; they enable the discovery of new relationships between variables and illuminate new pathways in behavioral analysis; and they address the endogeneity problem that arises from a causal relationship between the independent and dependent variables by using instrumental variables derived from lagged variables. In addition, this approach enables the estimation of consistent parameters, even when the time-series data are limited (Brañas-Garza et al. 2011; Trad et al. 2017).

This method was first introduced by Arellano and Bond (1991), and then, expanded upon by Arellano and Bover (1995) and Blundell and Bond (1998). For the sake of the analysis, the following two equations are estimated:

EVA_TA_it = α + β₁EVA_TA_i,t−1 + β₂X_i,t−j + u_i,t,  u_i,t = η_i + λ_t + ν_i,t

(4)

EVA_CE_it = α + β₁EVA_CE_i,t−1 + β₂X_i,t−j + u_i,t,  u_i,t = η_i + λ_t + ν_i,t

(5)

where subscript i denotes the ith sector (i = 1, …, N) and subscript t denotes the tth year (t = 1, …, T). EVA_TA and EVA_CE are the dependent variables with a one-year lag and X_i,t−j is the 1 × K vector of the current and lagged values of the additional explanatory variables. The error term u_i,t consists of the unobserved activity-specific effect η_i, the year-specific effect λ_t, and the disturbance term v_i,t, assumed independent across activities, whereby β₁ and β₂ are the parameters of interest. β₁ includes TL_EBITDA, DR, TAT, NPM, and HCE, while β₂ includes TL_EBITDA, DR, TAT, NPM, and HCE. To estimate Equations (4) and (5) the system GMM method (estimator) is applied.

The first step in this research is to analyze the data needed to evaluate the models before estimating them. The data from financial statements were collected for all non-financial entities operating in the Republic of Croatia from 2002 to 2021. The data collected over 20 years were structured into sixteen non-financial sectors according to national classifications of economic activities. Agriculture is an exception, as the analysis period includes periods from 2013 to 2020, considering the unavailability of data for all years. Other service activities under section S were also not analyzed due to the unavailability of data.

The population consisted of 61,674 entities in 2002, increasing to 137,436 entities in 2021. These entities employed between 745,000 and 920,000 employees, generating between EUR 52 and 113 billion in revenues, and between EUR 15 and 29 billion in value added.

The next step consisted of aggregating the data at the level of each non-financial activity. The dependent and independent variables were calculated using the aggregated inputs, resulting in a total of 320 observations. The calculations for the dependent and independent variables are presented in

Table 1. Financial ratios calculation.

Ratio	Ratio Calculation
Economic value added (EVA_TA)	Economic value added/Total assets
Economic value added (EVA_CE)	Economic value added/Capital employed (Capital and reserves)
Total liabilities to EBITDA (TL_EBITDA)	Total liabilities/Earnings before interests, taxes, depreciation and amortization
Debt ratio (DR)	Total liabilities/Total assets
Total assets turnover (TAT)	Total revenues/Total assets
Net profit margin (NPM)	(Net profit + Interest expenses)/Total revenues
Human capital efficiency (HCE)	Value added/Total employee costs (Human capital)

.

Since data for some years are missing, the final sample for both models consists of an unbalanced set of 320 annual observations of selected activity-level data.

The first panel model fits the economic value added/total assets ratio (EVA_TA) to the total liabilities/EBITDA ratio (TL_EBITDA), the debt ratio (DR), the total assets turnover (TAT), the net profit margin (NPM), and the human capital efficiency (HCE), while the second model fits the economic value added/capital employed ratio (EVA_CE) to the total liabilities/EBITDA ratio (TL_EBITDA), the debt ratio (DR), the total assets turnover (TAT), the net profit margin (NPM), and the human capital efficiency (HCE). Both models include one lag of the dependent variable (EVA_TA or EVA_CE) as explanatory variables and the TL/EBITDA ratio, the DR, the TAT, the NPM, and the HCE as regressors. To remove the cross-section’s fixed effects each variable in the regression is differenced. Concerning the dependent variable (EVA_TA or EVA_CE), the Arellano–Bond-type dynamic panel (predetermined) instruments include all valid lags, whereby a list of other instruments in the transformed equation consists of the TL/EBITDA ratio, the DR, the TAT, the NPM, and the HCE. The 2-step method is used to compute the Arellano–Bond 2-step estimator. The GMM weighting matrix employs White period weights, whereby the coefficient covariance method is based on ordinary estimates.

3. Results

The research results are presented in

Table 2. Descriptive statistics.

Value	EVA_TA	EVA_CE	TL_EBITDA	DR	TAT	NPM	HCE
Mean	−0.048255	−0.144724	13.00341	0.615075	0.645710	0.058465	1.708877
Median	−0.052485	−0.134679	9.431600	0.621439	0.571962	0.050913	1.592734
Maximum	0.085597	0.196929	192.3816	0.899399	1.558487	0.170711	4.539191
Minimum	−0.120784	−0.694351	1.107644	0.197862	0.127548	−0.165100	1.037068
Std. Dev.	0.033144	0.121359	15.44805	0.121553	0.371637	0.041472	0.472468
Observations	309	309	320	320	320	320	320

,

Table 3. Correlation coefficients.

Ratio	EVA_TA	EVA_CE	TL_EBITDA	DR	TAT	NPM	HCE
EVA_TA	1
EVA_CE	-	1
TL_EBITDA	−0.46	−0.42	1
DR	−0.06	−0.50	0.16	1
TAT	0.55	0.29	−0.38	0.29	1
NPM	0.45	0.40	−0.45	−0.27	−0.05	1
HCE	0.28	0.21	−0.13	−0.18	−0.23	0.51	1

,

Table 4. Dependent variable: EVA_TA.

Variable	Coefficient	Std. Error	t-Statistic	Prob.
EVA_TA (−1)	0.437748	0.162254	2.697922	0.0074
TL_EBITDA	−0.000325	0.000635	−0.511555	0.6094
DR	0.016556	0.044544	0.371683	0.7104
TAT	0.041332	0.058166	0.710587	0.4780
NPM	0.055967	0.131756	0.424778	0.6713
HCE	0.029717	0.015506	1.916494	0.0564
Effects Specification
Cross-section fixed (first differences)
Mean dependent var.	0.002089	S.D. dependent var.		0.019136
S.E. of regression	0.019611	Sum squared resid.		0.104222
J-statistic	15.29229	Instrument rank		16
Prob. (J-statistic)	0.121763
Arellano–Bond Serial Correlation Test
Test order	m-Statistic	rho	SE(rho)	Prob.
AR (1)	−2.786862	−0.041900	0.015035	0.0053
AR (2)	−0.699409	−0.006151	0.008795	0.4843

and

Table 5. Dependent variable: EVA_CE.

Variable	Coefficient	Std. Error	t-Statistic	Prob.
EVA_CE (−1)	0.326365	0.058931	5.538119	0.0000
TL_EBITDA	−0.002823	0.001762	−1.602598	0.1102
DR	−0.399010	0.098092	−4.067689	0.0001
TAT	0.027011	0.106635	0.253307	0.8002
NPM	0.370546	0.199631	1.856158	0.0645
HCE	0.052662	0.030913	1.703529	0.0896
Effects Specification
Cross-section fixed (first differences)
Mean dependent var.	0.004319	S.D. dependent var.		0.070271
S.E. of regression	0.082192	Sum squared resid.		1.830758
J-statistic	10.63779	Instrument rank		16
Prob. (J-statistic)	0.386424
Arellano–Bond Serial Correlation Test
Test order	m-Statistic	rho	SE(rho)	Prob.
AR (1)	−2.760629	−0.885203	0.320653	0.0058
AR (2)	1.252051	0.093191	0.074431	0.2106

. To gain a better insight into the characteristics of the data, descriptive statistics are calculated and presented in Table 2. The descriptive statistics include the calculation of the arithmetic mean, median, maximum and minimum values, standard deviation, and the number of observations.

It is visible that the mean values of EVA_TA and EVA_CE were negative during the observed period. There were 27 out of 309 cases where EVA was positive. A positive average EVA for 20 years was generated only by the information and communication sector. Although a positive EVA was generated in certain years in companies operating in human health and social work activities, education, mining, and quarrying, as well as in wholesale and retail trade activities, their average EVA for the 20 years was negative.

Standard deviation is a common measure of risk used in finance. A higher standard deviation indicates higher volatility, i.e., higher risk. From the results presented in the table, it is possible to notice that the lowest value of standard deviation was achieved by EVA_TA and NPM, indicating a lower degree of risk concentrated in these variables, while the highest value of standard deviation was achieved by TL_EBITDA, indicating a higher degree of risk concentrated in this variable.

From the table, it is also noticeable that the number of observations of selected activity-level data varies from 309 to 320. Since data from some years are missing, it is important to emphasize that our models will be based on unbalanced panel data. This is one of the limitations of this analysis regarding the reliability of the obtained results. Namely, models with balanced panel data provide less bias and higher reliability.

In the next step, the correlation coefficients between independent variables are calculated and presented in Table 3. The purpose of calculating correlation coefficients is to avoid potential negative issues with multicollinearity. Multicollinearity occurs if independent variables in a model are mutually correlated. The existence of multicollinearity will therefore result in less reliable statistical inferences.

The lowest standard deviation was achieved by EVA_TA and NPM and the highest by TL_EBITDA. To avoid potential problems with multicollinearity between variables, the correlation coefficients are estimated in the next step. Their values are presented in Table 3.

As stated by Kervin (1992), the absence of multicollinearity is achieved when the absolute value of correlation coefficients between variables is below 0.7. According to the results in Table 3, it is evident that the absolute values of the correlation coefficients are well below 0.7, indicating the absence of multicollinearity. The correlation coefficients confirm the existing research results on the direction of the relations between dependent and independent variables. Namely, debt ratios are negatively correlated with EVA-based ratios, while activity, profitability, and human capital efficiency ratios have positive correlations.

The results of the first model, where EVA_TA is the dependent variable, are presented in Table 4.

The GMM estimator demands the existence of first-order serial correlation, but no second-order correlation. Therefore, to test the serial correlation the Arellano–Bond test is applied. The models pass the first- and second-order serial correlation test. Furthermore, the J-statistic and its accompanying p-value indicate that over-identifying restrictions are set correctly. It is visible that the tests suggest model consistency. There are no signs of heteroscedasticity or autocorrelation. Insights from the t-statistics and their p-values indicate that only HCE is statistically significant, affecting EVA_TA positively, meaning that a rise in HCE increases EVA_TA.

The results of the second model, where EVA_CE represents the dependent variable, are shown in Table 5.

As with the first model, the second model also passes the Arellano–Bond serial correlation test. Over-identifying restrictions are likewise set correctly, as indicated by the J-statistic. The performed tests suggest model consistency without problems with heteroscedasticity or autocorrelation. Insights from the t-statistics and their p-values indicate that the DR, the NPM, and the HCE are statistically significant while the rest of the variables are insignificant. The DR negatively affects EVA_CE, meaning that a rise in the DR decreases EVA_CE, confirming the existing findings. Namely, a higher debt ratio (DR) generates higher costs of financing, resulting in lower EVA. On the other side, the NPM and the HCE positively affect EVA_CE, meaning that a rise in these ratios increases EVA_CE. An increase in NPM indicates higher profits per revenue unit, i.e., higher returns for companies that increase the EVA.

By comparing the results of the estimated models, it is possible to notice that HCE, as a segment of overall intellectual capital efficiency, which additionally includes structural and relational capital efficiency, is significant in both models and is positively contributing to generating EVA_TA and EVA_CE. Service sectors where intellectual capital is the central element of value creation generate higher EVA, so the companies are focused on investing in human capital which creates profit that exceeds the companies’ costs of capital.

The research results indicate that the number of significant variables in the second model is higher, while TL_EBITDA and the TAT are insignificant in both models. TL_EBITDA could be indicated as insignificant because of the variations in EBITDA among different sectors included in the analysis. Namely, some sectors are characterized as capital intensive, with higher levels of debt generating higher interest expenses and higher depreciation and amortization costs as well. TAT could vary among industries and years as well which could explain its insignificance in the derived models.

Finally, from the obtained results it is also noticeable that the impact of the DR on EVA_TA and EVA_CE varies in the estimated models. Specifically, in the first model, the impact of the DR on EVA_TA is positive, while in the second model, the impact of the DR on EVA_CE is negative. However, in the first model, the impact of the DR on EVA_TA is almost zero and it is statistically insignificant, while in the second model, the impact of the DR on EVA_CE is negative and statistically highly significant. The obtained results are not too surprising if we compare them with the correlation coefficients calculated and presented in Table 3. From Table 3 it is possible to see that both correlation coefficients between the DR and EVA_TA and EVA_CE are negative but the correlation coefficient between the DR and EVA_TA tends to be zero, which follows the obtained results of the estimated models.

4. Conclusions and Discussion

The main goal of the conducted research was to determine the impact of selected financial ratios on EVA-based ratios in Croatia. For this, two dynamic panel data models were estimated using the GMM (generalized method of moments) technique. Before that, the descriptive statistics of the variables included in the model were analyzed and showed that the mean values of EVA ratios were negative during the observed period. The lowest values of standard deviation were achieved by EVA_TA and NPM and the highest by TL_EBITDA.

The results obtained from the first GMM model indicate that only HCE is statistically significant in both models, affecting EVA_TA and EVA_CE positively, meaning that an increase in HCE generates increases in EVA_TA and EVA_CE and vice versa. While HCE is shown to be the only statistically significant variable in the first model, the obtained results from the second GMM model indicate that the DR and NPM are also statistically significant while the variables TL_EBITDA and TAT are insignificant. The DR negatively affects EVA_CE, meaning that a rise in the DR decreases EVA_CE and vice versa. On the other hand, the NPM positively affects EVA_CE, meaning that a rise in this ratio increases EVA_CE and vice versa.

Generally, it is possible to notice that HCE is significant in both models and positively affects EVA-based ratios. These ratios extend traditional ones, making a benchmark that can be useful for various stakeholders. It is also important to note that the second model includes a higher number of significant variables compared to the first model. However, it is worth mentioning that TL_EBITDA and TAT remain insignificant in both models. Ultimately, the paper answers the research question of which of the selected financial ratios impacts the EVA-based ratios. It could be concluded that HCE impacts both EVA_TA and EVA_CE, while the DR and NPM affect only EVA_CE.

While the EVA is commonly used as an absolute value indicator, its relation to total assets and capital employed has not been widely used. This paper suggests the introduction of two EVA-based ratios as a financial performance measurement tool that could become a benchmark for companies. The paper also indicates that human capital effectiveness is the variable that most significantly affects EVA-based ratios, so companies and policymakers should focus on increasing these ratios in the long run by strengthening human capital. The main challenge in increasing HCE lies in the fact that it could be achieved by decreasing the value of investments in human capital. Researchers and analysts should therefore pay attention to analyzing the directions of values that are used for calculating this ratio.

Finally, this study has some limitations. The first limitation is the cost of the capital calculation approach. The starting point for calculating the cost of capital of Croatian companies was from economies with different environments. Another limitation was the population. Namely, it included information from unaudited financial statements and information from the non-profit sector (but not budget users). The first limitation could be overcome by calculating the cost of capital more precisely for each activity, and the second by the inclusion of only profit sector entities with audited, and consequently more reliable, financial data. Another limitation consists of unbalanced panel data which gives lower precision results. Balancing the panel data could be achieved, but the sample would be smaller, which represents a limitation as well. In future research, the analysis could be extended using some additional financial ratios to broaden the existing findings. Regardless of the mentioned limitations, this analysis can serve as a solid foundation for future research on the impact of selected financial ratios on EVA-based ratios in Croatia.