1. INTRODUCTION

Intangible assets (IAs) have become an essential source of value for firms, especially in markets where networking, knowledge, and innovation are the new trade currency for high-growth companies (Andonova & Ruíz-Pava, 2016; Barth et al., 1998; Cardozo-Torres et al., 2021; Ciftci & Darrough, 2015; Dzenopoljac et al., 2017; Hall, B. H. et al., 2005, 2010; Lev & Daum, 2004).This has especially been true since the COVID-19 pandemic, when many firms understood that considerable investments in physical or tangible assets (TAs) would create impediments for companies to grow flexibly (De Nicola et al., 2021; Xiong et al., 2020).

IAs have been classified and defined in multiple ways in literature; however, there is no widely-accepted definition for their characteristics. Aside from this, ways of measuring them, and their impact on firm management, value, and performance are still under discussion (Barker et al., 2021).

It can be assumed that a high enough proportion of IAs should generate value for companies, yet the related scientific literature is not conclusive on this matter. Moreover, some results demonstrate a negative relationship between IAs and financial performance; in the case of Cardozo-Torres et al., several negative relationships between brand value and performance were reported (2021). Nevertheless, these divergences may be due to methodological differences exemplified by the companies chosen, how the data is collected, how IAs are measured, or even the financial system in which the measurement has been performed (Méndez-Morales & Yanes-Guerra, 2021).

The latter suggestion must be highlighted; some researchers measure IAs for companies traded on stock markets as the difference between market and book value; in such cases, IA value could be influenced by stock demand, causing prices to be volatile and creating either more or less Tobin's Q, goodwill or IA value. However, when this occurs, the relationship between performance and IAs goes from the accounting return to the value of market goodwill. Thus, the higher the return, the higher the demand for stocks, and the more significant the difference between book and market value (Haji & Ghazali, 2018). In those cases, higher IAs are possible whereas it is difficult to understand if this value is created by IAs or is influenced by market volatility.

In contrast, in other investigations, the value of IAs is based on how their accounts are represented, and there are discussions about how accounting standards allow companies to record their IAs (Barker et al., 2021). According to the International Financial Reporting Standards Foundation (IFRS), an IA is logged only when there is enough evidence of future related returns. Hence, under the IFRS, it can be assumed that a correlation between IAs and performance should arise, albeit in the long term, where returns are incurred after IAs are in the books.

Before the IFRS was created, countries used to apply their own accounting rules. Therefore, IAs were not only reported when the future return of IAs was proved but under firm accounting criteria. In the case of Colombia, our case study, pre-IFRS accounting standards did not underline the necessity of proving the future level of IA profitability (Andonova & Ruíz-Pava, 2016; Mesa Velásquez, 2012). Furthermore, most of the research about this topic has used data from publicly-listed companies because of the ease of access to their financial information, although these are not the norm in Colombia. Moreover, most research has been based on regions where companies tend to face less market stress and volatility, like the USA and Europe.

Despite the literature having many papers on the relationship between IAs and profitability, ours adds to it in various ways. Firstly, little research about this link has been performed in Colombia or the rest of Latin America (Andonova & Ruíz-Pava, 2016; Chiarelo et al., 2015); thus, analyzing Colombian company information using a panel data technique that has never been used to probe this relationship may shed new light on this topic. Secondly, we have found consistent evidence of a negative relationship between IAs and performance in the case of ROA (Return on Assets) and ROE (Return on Equity) for Colombian non-listed firms. Thirdly, this evidence is consistent with short- and long-term profitability measures. Fourthly, the relationship is consistent and robust to different measures of IA.

At the same time, the firms used in our sample have not floated on stock markets; this is noteworthy because a considerable part of the literature on this topic has focused on large corporations which are established and are based in developed countries; thus, they have reached a high enough performance level to allow them to trade on stock markets. It can be assumed that they are more financially sustainable than their smaller non-traded counterparts, implying that there is a selection bias in the literature by focusing on traded firms since it could influence the reported relationship between performance and IAs (Méndez-Morales & Yanes-Guerra, 2021).

Data used in this research came from Colombian firms before IFRS was adopted in 2016. For this research, data could have been used from after the foundation was implemented although the relationship between IAs and performance can only be analyzed in the long term and the global crisis created by COVID-19 must have generated a bias in this data; this is because companies suffered from downturns in profitability which would have led to an additional source of bias, so it was decided that a more extensive set of data should be used instead of one from the period 2016-2022.

Despite this, the data is limited bearing in mind that it comprises 4.89 years per company and does not disclose the age of the firms. Therefore, the sample does not allow us to distinguish the five-year average observation of an established firm from an entrepreneurial young firm of five years-old or less, in which case we would expect there to be a differential in the capacity to exploit IAs and thus limits the scope of the results.

2. LITERATURE REVIEW

3. METHODOLOGY

The data that we have used came from the Legiscomex1 database for the period 2005-2015. There was an unbalanced panel for 48,053 Colombian firms with an average of 234,952 observations and 4.89 years per firm. The Legiscomex database comprises companies' financial statements and calculated financial ratios. An unbalanced linear dynamic panel was used that is expressed as:

where (Y_it ) represents independent variables including return on assets (ROA), return on equity (ROE), net margin (NM), the EBIT2 margin (EBM), and the EBITDA3 margin (EBAM). Table 1 shows a summary of the profitability ratios used in this study and their correlations; all these ratios were calculated by the authors using final year figures and were winzorised at the 99% level. Appendix B shows the relationship between the variables used and past papers which adopted them.

Given that past profitability results of a company tend to reinforce contemporary profitability, a lag of return ratios ( $\delta Y_{i,t-1}$ ) was included to make this model dynamic; this variable expresses that firms receiving high returns in the past tend to receive high returns in the present; in addition, this variable helped us control for possible endogeneity issues in the model. An attempt was made to develop the generalized method of moments (GMM), including the first differences of variables to control for possible correlations with the error term. Unfortunately, the database had several data gaps, meaning that important information was missing; at the same time, there were no adequate instruments to perform the GMM methodology with different data other than lagged variables; thus, a dynamic panel model was selected with a lagged return variable as the prime methodology. Having data from more than forty thousand firms and ten years helped reduce the possibility of the error term being related to the lagged explained variable, as reported by Roodman (2009). Serpeninova et al. (2022), Skhvediani et al. (2022), Ge and Xu (2021), Cheikh and Noubbigh (2019), Ni, et al. (2020), Haji and Ghazali (2018), and Hartsema et al (2021) have previously used similar methodologies to confirm the relationship between IAs and returns.

In addition, independent control variables related to the financial behavior of the firm ( $x_{i,t}^j$ ) were included. The first one was the logarithm of total assets. This variable is also linked to firm size, affecting performance since one can assume that large firms tend to have better scale economies.

Next, the logarithm of operative revenues, a control variable that considers a firm to have higher profitability if it receives higher operative income, other aspects being equal. Besides this, the lag of non-operative revenues was included, given that, other things equal, the higher the non-operative revenues, the higher the firm's profitability. The debt-to-equity proportion was also included because it is assumed that highly leveraged firms tend to pay higher interest, as well as having higher tax shields, the higher the leverage is. Finally, the logarithm of non-operative revenues was also included because the higher they are, the higher the net income. Prior to the logarithmical transformation, a windsorization transformation was performed for every variable at 99%, owing to the fact that the variables had high outliers on the right side of the distribution.

A set of control parameters $Z_{i,t}^k$ was also included, formed of industry and localization dummies. The variable of interest in the model was $I{A}_{i,t-p}$ , measuring the logarithm of the book value of the IAs reported by the firms. In this case, the interest was in the full effect of book-activated IAs instead of individual IAs like trademarks, patents, or rights. Due to the effect of IAs not necessarily appearing in the short term, several models were run, including lags of the IAs; thus, the models were designed to shed light on the past inclusion of IAs and how they affected the profitability of the firm, so that four lags of IA variable were included in our study.

Fixed effects models were run where individual effects were given as ${\alpha }_i$ . A Hausman test based on the estimated disturbance variance from the efficient estimator was utilized to perform this model. Finally, models for heteroskedastic panels to correct standard errors and autocorrelation were employed; in every case, the results of the models were satisfactory.

4. RESULTS

To present the results of this research, two different approaches were used; firstly, tables 3 to 8 show the results for the econometric models (30 in total), including calculated betas, standard errors, level of significance, population and R square, among others.

Aside from this, several tables have been presented in Appendix C with abstract results of tables 3 to 8. In this appendix, the tables have been grouped by explanatory variables related to explained variables and the number of lags introduced in each model for the IA variable (from zero to four lags). Each model shows the explanatory variable having either a positive or negative relationship with the financial return variable and either a significant or a non-significant relationship with it.

For every explained variable, ROE, ROA, Gross Margin, EBIT margin, EBITDA margin, and Net margin, the results of their own lagged observations were negative and significant (see the first line of results in tables 3 to 8); This means that past profitability results negatively influenced contemporaneous returns. A possible explanation for this result is that firms cannot maintain positive increases in return rates indefinitely; therefore, a period of positive returns is likely to be followed by a decreasing period of profitability, whereas one with poor return results is likely to be followed by one with high returns, i.e., with performance being highly volatile over time.

Except for the ROE model (Table 3), profitability was positively related to the value of assets; thus, high investment in total assets creates returns for the company. This result seems logical because these firms tend to invest in all types of assets to increase their value. Since total assets are measured without including IAs, it can be said that physical and current assets tend to boost returns in Colombia's companies.

The results of ROE may be because of the decomposition of this ratio which used the DuPont analysis being negatively associated with the level of assets. On top of this, firms investing high values in assets may tend to do it using high equity levels, especially when it comes to SMEs, which have almost no access to credit, thus causing a decrease in the ROE ratio, given that equity increases in those cases.

As seems logical, an increase in revenues is related to higher profitability levels in ROE, ROA, EBIT, EBITDA, and Net margin (Tables 3, 4, 6, 7, and 8); all in all, a company with increasing sales tends to have high returns and good scale economies. Therefore, although it seems curious that gross margins tend to decrease with higher revenues, Colombian firms do not have those scale economies for production costs but they do have them for operating ones; however, more research is needed to confirm this.

In Tables 3, 4, and 8, non-operating revenues can be seen to be directly related to positive performance. Due to the fact that ROE, ROA, and Net margin are directly linked to net profitability, this result appears logical; gaining high values with other activities except for operating ones tends to increase the net profit of firms, increasing the return ratios using this value.

However, Tables 5, 6, and 7 show that non-operating revenues were negatively related to Gross, EBIT, and EBITDA margins. In the case of gross margin, the effect was significant for the models with zero, one, and two lags of IAs (models 11, 12, and 13). The increase in non-operating revenues seems to have diminished positive performance, at least for the operation. When a company deviates its focus from operating activities to non-operating ones, it appears that its operating performance tends to vanish.

As for the debt-to-equity ratio, in Tables 4, 5, 6, 7, and 8, one can see that capital structure is negatively related to performance. Thus, ROA, gross, EBIT, EBITDA, and Net margins tend to decrease if the company takes on more debt or diminishes the equity. Conversely, in Table 3, ROE is positively linked to debt structure. There is an intense discussion in financial literature about the effect of capital structure on performance; in several papers, negative relationships between both variables have been discovered, but in others, they have been positive (Abor, 2005; Ayaz et al., 2021). The fact that mixed relations have been spotted between performance and debt structure may strengthen the debate in the literature and be due to the diverse characteristics of firms in this sample.

IAs are negatively related to contemporaneous ROA and ROE; this means that the higher the IA levels, the lower the profitability ratios tend to be. Similar results can be seen in the literature (Cañibano et al., 2000; Chiao & Yang, 2011; Ge & Xu, 2021; Hartsema et al., 2021; Matias Gama et al., 2017; Qureshi & Siddiqui, 2021; Wahyuni et al., 2023). This result could be because of IA generation, book registration, and amortization causing a downturn in profitability in the short term. In addition, creating IAs is expensive and requires substantial investments in the firm.

Therefore, it could be assumed that increasing the number of IAs would create higher expenses for the firm. This would not only be because a firm would need to register some of the IAs with legal authorities but also because it would have to pay for appraisals, consultants, and other services to ensure that the registered fair value followed accounting rules.

In any case, Colombian companies need to amortize the value of IAs. The higher the amortization, the more the firm's profitability falls, even when this transaction does not affect cash flows. Since ROA and ROE are measured using net income, it seems logical for there to be a lower net ratio when firms increase amortized expenses.

One could think that if IAs created value within the firm, it would see an increase in revenue or a reduction in costs, using R&D, patents, or trademarks, which would tend to cancel out the negative effect of IA-related expenditure. However, this did not happen in the models; therefore, it can be said that the effect of IA-related expenditure is higher in the short term than a possible increase in revenue or a reduction in costs.

The effect of lagged IAs on ROE and ROA can be seen in Tables 3 and 4. One- and two- lagged IAs tend to be related to lower profitability. According to the previous explanation, the cost creation of IAs is higher than the profits obtained by the firm, even after two years. Although the amortization effect and the expenses related to the creation and registration of IAs may appear to explain this result, two years after registration, it does not seem logical that these expenses are the cause, whereas amortization does.

After the third lag, no relationship was found between IA, ROE and ROA. This result could be interpreted in two ways: firstly, companies may not obtain profits from IA registration in the short or the long term; secondly, the effect of expenses related to IA creation or registration seems to stop after the second year. It seems that even when accounting rules highlight that there should be proof that IAs will eventually create profitability, Colombian firms have IAs that do not create value for the firm.

Tables 5, 6, and 8 show that the book value of IAs was not related to gross, EBIT or net margin in contemporaneous or lagged periods. Thus, IAs did not affect all outcomes in the firm. More research needs to be done to understand why there are profitability ratios affected negatively by IAs while other margins seem to be independent of those effects.

Table 7 shows the effect of IAs on the EBITDA ratio. After two lags, there was found to be a positive and significant relationship between both variables, which seems to be an isolated case as no other positive effects of IAs on a firm's performance could be appreciated; therefore, it could be said that the effect was negative or non-existent in the short term.

5. ROBUSTNESS CHECKS

Given that we used an accounting measure of IAs, how we measured these may have affected the results obtained. Thus, by changing the measurement of IA variables in different ways, we performed several robustness checks.

Following the methodology of Andonova and Ruíz-Pava (2016), we included a proxy of IA ownership that took the value of 1 if the company had an IA value higher than zero (Own. dummy). 27% of companies had positive IAs (Table 9). We also incorporated a variable accumulating the value of the dummy variable for a specified number of years; companies had an average of 1.15 years of IA ownership with a maximum of 10 years, the total of the studied period.

Additionally, several models were run using other IA variations. Given that the original model used the log of capitalized IAs, we also included the ratio of IAs over total assets and revenue. Finally, the models were run using only companies with positive IA values, as a zero-inflated IA distribution could have affected the regression results.

In the case of the IA dummy, and its ratio over total assets and total revenue, the results were mostly the same as they were for the original models. IAs had a negative relationship with ROA and ROE and had no relationship with gross, net, or EBIT margins. EBITDA seemed to be affected positively but only after three years, with a 10% significance.

In the case of accumulated ownership, unlike the paper of Andonova and Ruíz-Pava (2016) we found no relationship in most of the models; however, we did note that after four years of ownership, IAs had a negative association with ROE and ROA, and a positive relationship with net margin after three years.

Including only positive values of IAs in the regression did not affect our results either; ROA and ROE had negative and significative relationships with IAs, but the gross, net, EBIT, and EBITDA margins were unaffected by IAs.

The results of these robustness checks seem to reaffirm the results of the original models in that IAs in in Colombian firms had either a negative or no relationship with performance when using accounting data. The results of these models are available upon request.

6. DISCUSSION

IAs are essential investments for companies worldwide; these assets tend to create high value, especially for firms in high technology sectors and for those where fixed physical assets do not affect business value. Most of the time, IAs are expected to create high returns in every firm, but according to the literature, it does not always do so.

Our results appear to concur with this. We believe that this evidence is related to the fact that Colombian firms, especially those not trading on stock markets, record IAs in books because they must do it to follow accounting rules, but in some cases, they use poor accounting practices for IA valuation. However, even when accounting rules require IAs to generate profitability, Colombian firms seem not to exploit them to create positive financial outcomes. According to Lev and Daum (2004), IAs by themselves do not create value. They should be combined with other production factors, and if they are used only for reporting purposes, they will negatively affect the firm's performance. A possible explanation for our results is that, on average, IAs in Colombian firms are capitalized but not exploited because they are not combined with other strategic investments to obtain positive results.

Another issue with IAs is that they create additional costs that seem substantial in the short and middle term. We could only see the benefits of IAs in the EBITDA ratio, and just for a specific period. By not using publicly traded firms in the used database, we were able to signal that for small, financially and operationally restricted firms, IAs did not create profits, or at least, any that would be traceable using accounting books. As may seem logical, such firms should gain some kind of benefit out of creating and maintaining IAs, although this should arise outside financial books.

On the other hand, the Colombian market may simply not be widely using IAs bearing in mind its poor positioning on international innovation indexes and the low investment that Colombian companies have made to create these assets (Cuellar et al., 2021; Méndez-Morales, 2019; Méndez-Morales & Yanes-Guerra, 2021). Therefore, even when a company develops and records IAs, the likelihood of its exploiting them to create profitability is low; IAs create additional costs that these firms cannot recover meaning that they should focus on exploiting them to create value for shareholders or to make profit.

The fact that we have found a clear negative relationship between ROE, ROA, and IAs makes it urgently needed for researchers in developing countries to study this phenomenon thoroughly. We do not believe that IAs are detrimental to a company per se; however, we need to understand how they can create value, going from accounting registration to value creation.

Finally, Governments in developing countries primarily focus on policies to enhance the creation of IAs, like patents and trademarks; nevertheless, our results have shown that the creation and capitalization of IAs is not enough to increase profitability; thus, governments should reinforce policies aimed at exploiting them rather than creating them.

7. CONCLUSIONS

We developed several models to determine if the book value of IAs was related to the profits of Colombian firms. Firstly, we used financial data for an unbalanced panel of 48,053 Colombian firms with an average of 234,952 observations and 4.89 years per firm.

After running more than 30 different dynamic panel models, we found that IAs negatively impacted ROE and ROA in the short and middle term and that the interest variable had no relationship with profitability ratios like gross, EBIT or net margins. Finally, we discovered a positive relationship between the EBITDA margin and IAs after two years; however, this was the only positive outcome in this research.

We have matched our results to two facts; firstly, Colombian companies do not exploit IAs to create profits, and these assets only create internal costs, which arise from the creation, appraisal and recording of IAs. Secondly, in Colombia, where there are poor results in innovation and low investments in R&D, the market and the possibility to exploit IAs is minimal. Finally, we used a set of firms that were not publicly-traded; consequently, these were more operationally and financially restricted than the floated ones used typically in literature.

The results of this study have opened new lines of research. Firstly, systematically comparing the differences between publicly and non-publicly-traded firms in several countries would be fascinating to unravel the effects of IAs. Secondly, given that our data from Colombia may be biased and its IAs market is rather limited, comparing our results against the ones in other developing countries and using accounting data would be interesting. Finally, more research needs to be done to understand why ratios like gross, EBIT and net margin have no relationship with the profits of Colombian firms.

8. LIMITATIONS

We selected a dynamic panel model to run our data. Therefore, we attempted to develop a generalized method of moments (GMM), including the first differences of variables to control for possible correlations with the error term. Unfortunately, our database had several data gaps, which meant that a considerable amount of important information was missing; at the same time, we did not have adequate instruments to perform the GMM methodology with data other than lagged variables; thus, we selected a dynamic panel model with a lagged return variable as our main methodology.

The data used in this research came from the financial records of Colombian firms. In most of the literature, the outcomes related to IAs were related to market capitalization. In this case, we used this data for a purpose: to aim to understand whether non-publicly-listed firms in a developing country behaved like the large publicly-traded firms in the literature; however, according to our results, non-publicly-listed Colombian firms did not obtain profitability from IAs like large publicly-traded firms in developed countries did.

Besides this, the database used had its limitations and the findings must be interpreted with caution for the following reasons: firstly, because the sample used had data for only 4.89 years per company on average, suggesting that companies over ten years old were rare in our sample, making it impossible to see the long-term effect of IAs in depth; secondly, we were unable to determine each company's age and could not contrast the results of the most recently created companies with the oldest ones as a result, preventing the findings of previous literature where companies could exploit intellectual property in the long term from being distinguished. In addition, the analysis did not include a measure of prior innovative capacity (e.g., number of patent applications), which could explain why some companies could obtain more significant financial value from some assets than others (e.g., patents), considering the limitations of innovative capacity.

Author contribution

Conceptualization, A.M.M., L.E.R.A. and C.A.M..; Methodology, A.M.M. and C.A.M..; Validation, A.M.M. and C.A.M.; Data Curation, L.E.R.A.; Formal Analysis, A.M.M. and C.A.M.; Funding Acquisition, A.M.M., L.E.R.A., and D.A.C.M.; Methodology, A.M.M. and C.A.M.; Writing – Original Draft Preparation, A.M.M., L.E.R.A., and D.A.C.M.; Writing – Review & Editing, A.M.M., L.E.R.A., and D.A.C.M. All authors have read and agreed to the published version of the manuscript.