This article seeks to complement the previous literature and clarify the particularities of the capital structure policy of firms with foreign direct investment in Angola.

This article seeks to identify the determinants of the capital structure of Portuguese firms with direct investment in Angola and to understand whether the determinants normally considered by standard finance theory are in line with those used by firms when structuring their capital structure policy to participate in the specific market of Angola.

This article examines 26 large Portuguese firms with investments in Angola using econometric panel data for the period 2006–2010.

The study applied fixed and random effects methods and panel-corrected standard errors that maintain efficiency and unbiased behaviour even in the presence of panel-level heteroscedasticity and contemporaneous correlation of observations among panels.

The results provide evidence that the determinants normally considered by standard finance theory are in fact – in terms of sign and coefficient dimension – those used by firms for structuring their capital structure policy when involved in the internationalisation process of entering Angola. Specifically, age, asset structure, return on assets and tangibility have a positive influence on the capital structure of Portuguese firms that have invested in Angola, while non-debt tax shields and liquidity have a negative influence on these companies’ leverage ratios. When comparing our results with studies that have analysed the capital structure determinants of listed Portuguese firms – firms belonging to the PSI 20 Index and large firms in the Portuguese corporate sector – we found similarities in the sign and coefficient dimension of the determinants of capital structure. However, the profitability coefficient sign is in line with the trade-off framework (i.e. profitability is positively related to debt) but not with pecking order theory (i.e. profitability is negatively related to debt).

Our results suggest that the high-growth Angolan market is seen by larger Portuguese firms as a low-risk diversification process because of the economic hardship Portugal has gone through, as well as cultural and linguistic similarities to Portugal. As such, the Angolan market is seen as an extension of the Portuguese domestic market that has increased potential. This scenario potentially reduces the firm default probability and the cost of debt. Maintaining the tax shield benefits of debt and decreasing the cost of debt – through a reduction in the default probability – have induced profitable firms to use more debt.

Capital structure has always been a polemical subject when it comes to financial theory. The selection of which capital structure to follow in a given project is a crucial decision, given the need to maximise returns for all firms’ stakeholders (Abor & Biekpe

Since the groundwork research done by Modigliani and Miller (

Although much research has been done on market imperfections, bankruptcy costs and information asymmetry, there are few studies on firms’ capital structure in Africa (Abor

Angola had a civil war from 1975 to 2002. The end of the civil war provided a wide range of investment opportunities, mainly because of the shortage of numerous goods and services, as well as the presence of a significant domestic market with a high purchasing power in some social strata. This opportunity combined with the small psychic distance between Portugal and Angola to open a window of opportunity for many Portuguese firms to invest in Angola, where they innovated and sometimes created completely new concepts for this market but, sometimes, took on high risks. By studying the decisions made about the capital structure of Portuguese firms with investments in Angola, one can explore how these investors decide what the best capital structure policy for their investments is, providing this field of study with a new perspective.

This article analyses Portuguese firms that have direct investments in Angola, in the form of either branches or subsidiaries but with headquarters in Portugal, with the goal of identifying the determinants of their capital structure. The objective is to understand whether the determinants normally considered by standard finance theory are, in fact, those used by firms for structuring their capital structure policy when participating in the specific market of Angola.

The article is organised as follows: After this brief introduction, a review of the literature on capital structure is described in the section ‘Capital structure: Literature review’. In the section ’Determinants of capital structure and research hypotheses’, the determinants of capital structure are examined. In the section ‘Data, estimation methodology and model’, the research methodology is presented. The results are presented in the section ‘Main results and discussion’. Lastly, the discussion of findings and conclusions are presented.

The capital structure theory has been quite often debated in the corporate finance literature. It concerns the ways firms use equity and debt capital to finance their assets.

Modigliani and Miller (

Myers (

Brealey, Myers and Allen (

According to Crnigoj and Mramor (

According to the pecking order theory, firms do not seek an optimal capital structure. The structure instead reflects financing options taken in the past (Myers

This information asymmetry theory also is one of the fundamental theories that help explain capital structure. Information asymmetry occurs when managers have more information than investors (Brealey et al.

According to Istaitieh and Rodríguez-Fernández (

Regarding the second theory, the level of debt might affect the market, and the market structure also can impact the capital structure of firms (Santos, Moreira & Vieira

In their definition of capital structure, Silva and Queirós (

Boateng (

In addition, Parsons and Titman (

Titman and Wessels (

Hovakimian, Hovakimian and Tehranian (

Rocca et al. (

Saito and Hiramoto’s (

As leverage is negatively influenced by asset structure, firms tend to prefer short-term rather than long-term debt (Daskalakis & Psillaki

Brito, Corrar and Batistella (

Age is usually expected to have a positive effect on capital structure – as represented by debt to equity ratio – given that firms increase their liquidity capacity over the years. Abor and Biekpe (

For Bhaird and Lucey (

Liquidity is measured through the current assets/current liabilities ratio, providing information on whether firms can meet their short-term financial commitments. Ahmed et al. (

The level of intangible activity is expected to have a positive impact on leverage as firms with high expenditures on research and development need higher levels of external capital than internal capital. This result suggests that internal financing is not enough to support the high-level growth of these firms regarding their increased need for investment (Bhaird & Lucey

Based on pecking order theory, Myers (

Some other studies (Chadha & Sharma

The pecking order theory is especially appropriate for small and medium-sized firms … These firms do not typically aim at a target debt ratio … Instead, their financing decisions follow a hierarchy, with a preference for internal over external finance, and for debt over equity. (pp. 325–326)

Abor and Biekpe (

Firms with higher tangible assets can use debt more easily as creditors believe these firms can fulfil their obligations more easily. Therefore, tangibility should positively influence leverage.

While Couto and Ferreira (^{1}

Non-debt tax shields (NDTS) are characterised by the weight assigned to the depreciation of assets (Rebelo

The growth of firms is expected to have a positive impact on leverage. However, some studies have concluded otherwise, which might be a consequence of using different variables for growth, such as net assets growth rate (Couto & Ferreira

Another definition given for growth is firms’ market value divided by the firms’ book value (Karadeniz et al.

Studies have produced contradictory results. While Couto and Ferreira (

Karadeniz et al. (

Lastly, Abor and Biekpe (

As large firms are usually more diversified than smaller firms are, they are less prone to financial difficulties and have fewer bankruptcy costs (Brito et al. ^{2}

Although Nunkoo and Boateng (

Bhaird and Lucey (

Studies of capital structure and the main factors considered when choosing it are increasingly important in the corporate finance literature. In this study, the goal was to analyse a specific case not studied before: Portuguese firms with direct investments in Angola. Thus, this study had the objective of verifying how the following factors influence the capital structure of Portuguese firms with subsidiaries or branches in Angola: asset structure, age, liquidity, intangibility, profitability, tangibility and NDTS. Although other determinants influence capital structure (i.e. growth and size), they were not included in this study because of the unavailability of data for the firms in our sample.

Explained variables, explanatory variables and hypotheses.

Variables | Hypotheses | Variables description |
---|---|---|

Leverage ratio | - | Leverage_Ratio = Ln(total liabilities/total assets) |

Asset structure | H1: A relationship exists between capital structure and asset structure | Asset_Structure = Ln(fixed assets/total assets) |

Age | H2: A relationship exists between capital structure and age | Age = Ln(Age) |

Liquidity | H3: A relationship exists between capital structure and liquidity | Liquidity = Ln(total net assets/short-term debt) |

Intangibility | H4: A relationship exists between capital structure and intangibility | Intangibility = Ln( intangible assets) |

Profitability | H5: A relationship exists between capital structure and profitability | Return_on_sales = Ln(net income/sales); Return_on_Assets = Ln(earnings before interests/total assets) |

Tangibility | H6: A relationship exists between capital structure and tangibility | Tangi = Ln(tangible fixed assets/total assets) |

Non-debt tax shields | H7: A relationship exists between capital structure and NDTS | Non-Debt_Tax_Shields = Ln(depreciation/total net assets) |

The sample in study comprises 26 firms dating to the period 2006–2010, which were chosen according to the availability of data. The data were gathered by using a list of the Portuguese firms with direct investments abroad, provided by the Agency for Investment and External Trade of Portugal in Luanda.

Based on this list, an analysis of the firms’ websites was conducted with the objective of collecting the information available. Firms without any website or information were first contacted via e-mail to acquire the required data. In the cases in which no information was acquired, telephone calls were made. The banks on the list were not considered, given that they have a different tax code, accounting rules and operation modes. Information, thus, was collected from the following firms: Compta, Auto Sueco, FDO, Conduril, Enoport, Eurico Ferreira, Galp, Visabeira, Martifer, Mota-Engil, Monteadriano, Orey, Petrotec, PT, Sumol, Tomás de Oliveira, Obrecol, EFACEC, Glintt, Opway, M. Couto Alves, Abrantina, Soares da Costa, Somague, Teixeira Duarte and MSF.

The data were extracted from the consolidated financial reports and statements of these 26 firms, which were normally made available on their websites. The collected data cover the following variables: total assets, fixed assets, tangible fixed assets, intangible assets, total net assets, depreciations, equity capital, short-term debt, long-term debt, total debt, liabilities, net profits, distributed dividends, sales, earnings before taxes, price per share, size and age.

According to Hsiao (

Fixed and random effects models were considered (two estimation methods inside the panel data models) when specifying the econometric model. The rationale behind the choice between the two models are: the fixed effects model is the most suited to analyse the exclusive impact of variables that change over time – that is, this model is suitable for studying the causes of change inside an entity (Gujarati

Complementarily, a test created by Hausman allows researchers to ascertain which model is more suitable: the null hypothesis assumes that the random effects estimator is the most appropriate (Johnston & Dinardo

Although, panel data models can be estimated even when there are severe deviations from the classical assumptions and ‘complex error compositions’ are present (Basu & Rajeev

Therefore, the panel data model needs to be complemented by carrying out tests to verify the presence or absence of

The procedure needs to be as follows. After testing for the presence of heteroscedasticity and panel autocorrelation, if any deviation from the classical assumptions is detected, Beck and Katz’s (^{3}^{4}

Marques and Fuinhas (^{5}

To carry out the data analysis in this study, static panel data and econometric methodologies using the program STATA 11 were chosen. This study’s data produced an unbalanced panel, given the lack of information for all variables in all the years covered. ^{6}

Following established procedure, an initial analysis of the data was made. The results of the specification tests are outlined in

Specification tests

Type of test | Random effects | Fixed effects |
---|---|---|

Modified Wald test (χ2) | - | 234.59 |

Wooldridge test F(N(0,1)) | 4.501 | 4.501 |

Hausman test | 20.26 |
- |

shows a significance level of 1%.

A Wooldridge test was carried out to test the presence of autocorrelation. The results support the following conclusion: at a significance level of 5%, the null hypothesis of no first order serial correlation cannot be rejected, that is, there is no serial correlation.

Following Baum’s (

Given that there is a deviation in classical assumptions, in particular when it comes to the existence of heteroscedasticity, the PCSE estimator had to be used to rectify the deviation.

Models with application of the robust standard error to random and fixed effects models

Variables | PCSE |
RSE |
CSE |
|||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|

Hetonly |
Corr(AR1) hetonly |
Fixed effects |
Random effects |
Fixed effects |
Random effects |
|||||||

Coefficient | Coefficient | Coefficient | Coefficient | Coefficient | Coefficient | |||||||

Asset_Structure | 0.2079 | 0.009 |
0.2011 | 0.016 |
0.1703 | 0.316 | 0.2174 | 0.098 |
0.1703 | 0.050 |
0.2174 | 0.003 |

Age | 0.1483 | 0.000 |
0.1390 | 0.017 |
0.2784 | 0.267 | 0.1473 | 0.052 |
0.2784 | 0.400 | 0.1473 | 0.046 |

Liquidity | -0.4566 | 0.000 |
-0.5174 | 0.000 |
-0.6412 | 0.000 |
-0.5255 | 0.007 |
-0.6411 | 0.000 |
-0.5255 | 0.000 |

Intangibility | 0.0148 | 0.289 | 0.0149 | 0.341 | -0.0205 | 0.271 | 0.0128 | 0.508 | -0.0205 | 0.542 | 0.0128 | 0.479 |

Return_on_sales | 0.0248 | 0.378 | 0.0291 | 0.318 | 0.0382 | 0.353 | 0.0322 | 0.238 | 0.0382 | 0.315 | 0.0322 | 0.210 |

Return_on_assets | 0.1178 | 0.073 |
0.1517 | 0.046 |
0.4089 | 0.008 |
0.1780 | 0.175 | 0.4089 | 0.000 |
0.1780 | 0.002 |

Tangi | 0.1675 | 0.029 |
0.1997 | 0.031 |
-0.1126 | 0.498 | 0.2158 | 0.091 |
-0.1126 | 0.623 | 0.2158 | 0.028 |

Non-Debt_Tax_Shields | -0.1999 | 0.015 |
-0.2664 | 0.003 |
-0.2093 | 0.271 | -0.2586 | 0.083 |
-0.2093 | 0.021 |
-0.2586 | 0.000 |

Constant | -1.1304 | 0.002 |
-1.2423 | 0.006 |
-1.6314 | 0.081 |
-1.1820 | 0.002 |
-1.6314 | 0.223 | -1.1820 | 0.023 |

Observations | 70 | - | 70 | - | 70 | - | 70 | - | 70 | - | 70 | - |

^{2}/Pseudo ^{2} |
0.5717 | - | 0.5913 | - | - | - | - | - | - | - | - | - |

F test | - | - | - | - | 12.29 | 0.000 |
- | - | 15.72 | 0.000 |
98.11 | 0.000 |

Wald (χ2) | 82.69 | 0.000 |
57.82 | 0.000 |
- | - | 37.49 | 0.000 |
0.1703 | 0.050 |
0.2174 | 0.003 |

PCSE, panel-corrected standard errors; RSE, robust standard errors; CSE, conventional standard errors.

Dependent variable: Liabilities/Total Assets; Corr(AR1) – first order autocorrelation AR(1), in which the coefficient of AR(1) is the same for all panels; hetonly specifies that deviations are taken as heteroscedastic. The item of test F tests the null hypothesis of there not being significance for the model on the whole (of the estimated parameters). The Wald test (χ2) evaluates the null hypothesis of there not being significant for all coefficients of all explanatory variables.

***, **, *, refer to significance level from 1%, 5% to 10%, respectively.

The results of the fixed effect and random effect models, with CSE and RSE, allow a comparison of the robustness of results achieved by the PCSE estimator with the results achieved by fixed and variable effects estimators. The analysis was then continued using data from the PCSE estimator as it is considered to be more robust and to provide better results.

From all the hypotheses subjected to empirical analysis, only two were not statistically significant: Hypotheses 4 and, partially, 5. These postulated the existence of a relationship between capital structure and the level of intangibility and return on sales and/or return on assets. For the second hypothesis, only the relationship with return on sales was not statistically significant.

The analysis results shown in

As can be observed, the asset structure (i.e. Hypothesis 1) has a positive impact on the leverage ratio (0.2079). This conclusion goes against what Rebelo (

A firm’s age (i.e. Hypothesis 2) also has a positive impact on the leverage ratio (0.1483). This conclusion agrees with the results achieved by Abor and Biekpe (

Liquidity (i.e. Hypothesis 3) has a negative impact on the leverage ratio (-0.4566), which is in agreement with the results obtained by Sbeiti (

Profitability was divided into return on sales and return on assets (i.e. Hypothesis 5). However, only the positive (0.1178) relationship between the leverage ratio and returns on assets is statistically significant, which is in agreement with Psillaki and Daskalakis (

The assets’ tangibility (i.e. Hypothesis 6) has a positive influence on the leverage ratio (0.1675) since firms with higher tangible assets can more easily use debt because they have collateral to present to banks. This conclusion was also reached by Nunkoo and Boateng (

NDTS (i.e. Hypothesis 7) maintain a negative relationship with the leverage ratio (-0.1999). This conclusion contradicts the findings of authors such as Rebelo (

In this study, the factors that influence the choice of the capital structure of 26 Portuguese firms with investments in Angola were examined. With the exception of these firms’ profitability, we found similarities in the sign and coefficient dimension of capital structure determinants when we compared our results with studies that analysed the capital structure determinants of listed Portuguese firms, firms belonging to the PSI 20 Index and large firms in the Portuguese corporate sector (Antão & Bonfim

Regarding the results, as expected, asset structure is positively related with the leverage ratio, with a coefficient of 0.2079. Likewise, tangibility is also positively related with the leverage ratio. When the sign and coefficient dimension of the relation between tangibility and leverage in our study is compared to the above-mentioned studies, we can conclude that our result of 0.1675 is similar to the other studies’ results. Serrasqueiro and Rogão (

A firm’s age and profitability – represented by the return on assets – are also positively related with the leverage ratio. In the specific case of the relationship between profitability and leverage, compared with other studies in terms of sign and coefficient dimension, our result of 0.1178 has the same dimension but with an inverse sign in contrast to what was obtained by Serrasqueiro and Rogão (

Lastly, liquidity and NDTS are negatively related with the leverage ratio. The liquidity coefficient is -0.4566, as compared with Antão and Bonfim’s (

In general, this leads to the conclusion that the capital structure determinants normally considered by standard finance theory are, in fact – in terms of sign and coefficient dimension – similar to those used by Portuguese firms investing in the Angolan market. However, the sign of the profitability coefficient is in line with the trade-off framework (i.e. profitability is positively related to debt) and not with pecking order theory (i.e. profitability is negatively related to debt). The explanation offered for this finding is that internationalisation to Angola is seen by Portuguese firms as a diversification strategy involving a market in which the cultural and language differences are quite low. As such, taking into account the high-growth rate of the Angolan market

It is possible to advance that the determinants of the capital structure of firms from less intensely developed countries investing abroad on the African continent do not necessarily differ from other previous studies (Abor

This research also has important implications as it complements studies especially in less-endowed countries or in Africa. For example, results obtained in the hospitality industry in Turkey (Karadeniz et al.

It is important also to emphasise that foreign firms investing in Africa need to seriously take into account corporate governance decisions as government policies may not only discriminate between indigenous and foreign shareholders (Boateng

This study has some limitations that conditioned the research. The first limitation is the small dimension of the sample, which consisted of 26 firms. Another limitation is the size of the firms studied. As they are all large firms, it was not possible to test whether size influences capital structure. This limitation is linked to the ease of obtaining data from large firms, as opposed to smaller firms. Lastly, not all firms had data for every year considered in this study. Although this lack of data is situational, it led to an unbalanced panel.

Considering the results of this study, it would be interesting to study in the future the same factors in firms of different sizes, such as small- and medium-sized businesses versus large firms. The purpose of this future study would be to discover the strategic differences between both types of firms.

The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.

The article is a joint work of the two authors in all of its phases.

Summary statistics.

Variable | Observations | Mean | Standard deviation | Minimum | Maximum |
---|---|---|---|---|---|

Leverage_ratio | 104 | 1.099187 | 3.391286 | 0.1881 | 35.2465 |

Asset_structure | 104 | 0.3917452 | 0.2297208 | 0.0093 | 0.8718 |

Age | 102 | 47.97059 | 30.07629 | 1 | 110 |

Liquidity | 104 | 2.22868 | 6.71436 | 0.0019 | 50.2408 |

Intangible_activity_level | 100 | 6.73E+08 | 2.06E+09 | 0 | 1.02E+10 |

Return_on_sales | 88 | 4.648626 | 22.63219 | -4.9775 | 179.7684 |

Return_on_assets | 104 | 2.504371 | 2.539035 | -0.0243 | 11.5703 |

Tangibility | 104 | 0.1914587 | 0.113603 | 0.014 | 0.4168 |

Non-debt_tax_shields | 104 | 0.4322817 | 3.817053 | 0 | 38.9781 |

For more discussion see among others Serrasqueiro and Rogão (

For more discussion see, among others, Couto and Ferreira (

Which is the most efficient when the data do not have serial correlation (autocorrelation), comparing its results with those achieved from the classical panel data estimators (Fixed and Random Effects).

Additionally, it allows: the error term to be correlated over the firms, the use of firstorder autoregressive process for error term over time and the error term to be heteroscedastic (Cameron & Triverdi,

In this model, β values represent the coefficients of the independent variables. _{it}represents the error term.

From observation of the descriptive statistics, it is possible to conclude that the debt of Portuguese firms with direct investment in Angola is on average around 110%. Moreover, the liquidity, return on sales, return on assets and NDTS are variables with some degree of volatility, as their standard deviations are above their mean values. The Portuguese firms with direct investment in Angola appear to be moderately volatile in asset structure, intangibility and tangibility, which suggests some degree of stability.