분위회귀분석을 이용한 종합건설회사의 자본구조 분위별 결정요인에 관한 실증적 분석

Abstract

본 연구는 1997년 IMF 외환위기, 2008년 국제금융위기, 유럽재정위기와 같은 반복적인 거시경제 위기에서 종합건설회사의 반복적인 부도의 원인을 과도한 타인자본의 활용과 같은 자본구조에 있다고 판단하였다. 이를 기반으로 자본구조와 관련된 이론과 선행연구를 고찰하였으며, 실증모형 구축을 위한 연구방법론 또한 알아보았다. 관련 이론 및 선행연구 고찰 결과 건설산업과 관련된 연구는 매우 제한적이며, 자본구조 결정요인에 대한 일관된 견해는 찾을 수 없었다. 뿐만 아니라, 건설회사의 과대 또는 과소 차입 자본구조가 어떠한 결정요인에 의해 결정되는지를 설명할 수 있는 실증분석방법은 과거 선행연구에서도 찾아보기 어려웠다. 과거 선행연구는 자본구조 결정요인에 대한 실증과 목적집단간 결정요인의 차이점을 밝혀나가는데 중심을 두고 있었으며, 최근 국내외 다른 산업에서 분위회귀모형을 이용한 자본구조 결정요인의 변화행태와 비선형성에 대한 연구가 이루어지고 있었다. 이에 따라 본 연구는 2000년부터 2013년까지 국내 상장·외부감사대상 종합건설회사를 대상으로 분위별 자본구조 결정요인에 대한 실증분석을 수행하였다. 실증분석에 사용된 표본의 수는 대형 건설회사 35개, 중소형 건설회사 196개로 총 231개의 건설회사 재무정보를 활용하였다. 실증분석의 목적은 종합건설회사의 자본구조 결정요인의 행태(부호성)가 기존 이론과의 괴리 및 유사성이 존재하는지를 검증하는 것과, 종합건설회사의 과대 또는 과소 자본구조 분위에 따라 결정요인이 변화하는 행태 및 비선형성이 존재하는 지를 검증하는 것, 표본집단 간 자본구조 결정요인의 변화행태가 서로 차이를 보이는 지를 검증하는 것이다. 위와 같은 연구의 목적을 달성하기 위해 각각의 연구가설을 설정하였으며, 합동최소자승모형, 일반화최소자승모형, 분위회귀모형을 활용하여 자본구조 결정모형을 구축한 뒤 가설을 검증하였다. 실증분석 결과는 다음과 같이 세 가지로 요약될 수 있다. 첫 번째, 건설회사의 자본구조 결정요인의 행태(부호성)가 기존이론과 괴리 및 유사성을 가지고 있는가를 검증한 결과, 종합건설회사의 자본구조 결정요인은 정태적 상충이론과 자금조달 순위이론 모두 부분적으로 지지하는 것으로 나타났다. 이는 곧 종합건설회사는 기업의 규모적 이점을 가지고 부채를 활용하는 패턴을 보이지만, 경우에 따라 내부자금을 활용하여 부채를 적게 사용하는 패턴도 함께 나타냄을 의미한다. 가설검증 결과에 비추어 종합건설회사의 자본구조를 효과적으로 설명해주는 결정요인은 기업의 규모(+), 수익성(-), 유동성(-) 변수로 표본집단별로 대부분의 자본구조 분위에서 통계적으로 유의하며 일관성 있는 부호성을 나타내었다. 수익성과 유동성 변수의 추정계수 값이 크다는 것은 건설회사의 자본구조 결정이 자금조달 순위이론을 지지한다는 의미와 같은 맥락을 가지고 있다. 하지만, 기업의 규모나 비부채감세효과(평균 분위 이하에서) 또한 강건한 영향을 주고 있음에 따라 정태적 상충이론도 부분적으로 지지하고 있는 것으로 분석되었다. 두 번째, 종합건설회사의 과대 또는 과소 자본구조 분위에 따라 결정요인이 변화하는 행태 및 비선형성이 존재하는 지를 검증한 결과, 기업의 규모, 수익성, 유동성 변수가 대부분의 자본구조 분위에서 강건한 부호성을 가지면서 변수의 영향이 변화하는 행태를 나타내었다. 비부채감세효과 변수는 평균 차입 분위 이하에서만 통계적인 유의성을 확보할 수 있다 하더라도 회귀계수의 변화가 관련 이론 및 가설에 부합함에 따라 건설회사의 자본구조를 설명하는 요인으로 활용가치는 있는 것으로 분석되었다. 결과적으로 종합건설회사의 자본구조 분위별 결정요인은 자본구조 분위에 따라 변화한다는 가설을 부분적으로 채택할 수 있었다. 이는 전통적인 자본구조 결정모형 추정방식의 조건부 평균에 추정계수가 집중되는 현상을 배제하고 그 동안 과대 혹은 과소 추정되어왔던 새로운 영향치를 식별할 수 있는 계기가 되었다. 세 번째, 표본집단 간 자본구조 결정요인의 변화행태가 서로 차이를 보이는 지를 검증한 결과, 0.1 분위에서 비부채감세효과가 표본집단별 회귀계수 차이 검증에서 차이를 나타내었다. 그리고 0.6 분위에서는 자산고정성 변수가 집단 간의 차이를 나타내었다. 결론적으로 표본집단별로 자본구조 결정요인에 차이가 존재할 것이라는 가설을 매우 국소적인 부분에서 채택하였다. 이와 같은 결과는 표본집단 간 자본구조 결정요인이 거의 유사하다는 것을 의미한다. 따라서 앞서 연구의 배경에서 제시된 기업의 규모 차이에 따르는 자본구조 결정행태의 차이는 표본집단내에 존재하지 않는 것으로 분석되었다. 실증분석 결과 기업의 경영관점, 정부 정책적 관점, 연구방법론의 관점에서 다양한 시사점을 제시하였다. 기업의 경영관점에서는 대형 또는 중소형 건설회사의 자본구조 결정모형을 추정하여 건설회사의 자본구조 결정에 대한 의사결정 체계를 식별할 수 있었다는데 있다. 뿐만 아니라 과소 차입에서부터 과대 차입 분위까지의 의사결정 체계의 스펙트럼(spectrum)을 제시함으로써 개별 건설회사 스스로 회사의 자본구조 상황에 빗대어 의사결정을 할 수 있게 도움으로써 기존의 연구와는 차별화된 결과로 해석할 수 있다. 정부 정책적 관점에서는 자본구조 분위별 대형, 중소형 건설회사의 경영패턴을 분석하는데 기초적인 정보로 활용할 수 있을 것으로 판단된다. 과거의 일반화된 자본구조 결정행태만으로는 건설회사의 자본구조 분위별 경영패턴을 파악할 수 없음에 따라, 각종 경영패턴에 맞는 정책을 개발하는데 어려움이 존재하였다. 본 실증연구 결과는 건설회사의 서로 다른 경영패턴(자본구조 의사결정)을 제시함에 따라, 보다 현실적이며 구체적인 정책개발에 도움을 줄 것이다. 그리고 대부분의 육성/금융 정책들이 자금조달에 어려움을 겪고 있는 중소기업에 집중되어있는데, 본 연구결과에서 도출된 중소기업의 분위별 자본구조 결정요인을 활용할 경우 어떠한 환경을 조성해 주어야하는지, 어떤 부분에서 지원을 해주어야 중소기업의 자금조달환경을 완화시켜줄 수 있는지를 파악할 수 있을 것으로 판단된다. 연구방법론 관점에서 도출된 시사점은 먼저 전통적 추정방법을 이용한 자본구조 결정모형은 과대, 과소 차입 경영전략을 구사하는 기업의 자본구조 의사결정을 파악할 수 없다. 이에 따라 본 연구에서는 다양한 가설 검증을 통해 분위회귀모형의 추정결과가 앞선 전통적 모형의 추정결과보다 더 다양하게 도출된다는 것을 확인할 수 있었다. 이와 같은 결과가 분위회귀모형이 다른 분석방법에 비해 절대적으로 우수하다는 결론을 내릴 수 없으며, 두 모형을 적절히 조합하여 활용하는 것이 보다 객관성 있는 연구 결과를 도출할 수 있을 것을 의미한다.|This study judged that repeated bankruptcy of general contractors was caused by their capital structure using excessive borrowed capital under the circumstances of repeated macroeconomic crisis such as IMF foreign exchange crisis in 1997, international financial crisis(2008) and Europe’s financial crisis(2009). Based on the foregoing judgment, this study considered theories and previous researches related to capital structure, and also looked into a research methodology for establishing an empirical model. As a result of consideration of related theories and preceding researches, this study found that researches related to construction industry were very limited and failed in finding a consistent opinion on capital structure determinants. Moreover, it was hard to find a method of empirical analysis in previous researches explaining by what determinants construction company’s over or under borrowing capital structure is decided. Preceding researches were giving priority to substantiation of capital structure determinants and investigation of differences in determinants between sample groups, and researches on changed behavior and non-linearity of capital structure determinants using a quantile regression model are being carried out in other domestic and foreign industries in recent days. Accordingly, this study conducted an empirical analysis on the capital structure determinants by quantile targeting domestic general contractors that are listed·externally audited from 2000 through 2013. Samples used in the empirical analysis include 35 large construction companies, 196 small and medium construction companies, and this study utilized financial information of total 231 construction companies. Objectives of empirical analysis were to verify whether or not gap and similarity exist between behavior (sign) of general contractor’s capital structure determinants and existing theories, verify if changing behavior and non-linearity exist in determinants depending on general contractor’s over or under capital structure quantile and verify whether or not changed behavior of capital structure determinants shows difference between sample groups. To achieve the above study objectives, this study established research hypotheses respectively, and verified the hypotheses after establishing a capital structure decision model by using pooled-ordinary least squares model, generalized least squares model and quantile regression model. Results of empirical analysis can be summarized as following three. First, as a result of verifying whether or not gap and similarity exist between behavior (definite) of construction company’s capital structure determinants and existing theories, it was proved that general contractor’s capital structure determinants support both static trade-off theory and peaking order theory in part. This means that general contractors show a pattern using debt with a advantage of company’ size, but also shows a pattern that uses debt on a shoestring by utilizing internal capital as circumstances require. In the results of hypothesis verification, determinants that explains general contractor’s capital structure effectively are the variables of company’s scale (+), profitability (-), liquidity (-), and each sample group showed a statistically significant and consistent sign in most capital structure quantiles. Meaning that the value of estimated coefficient is high in variables of profitability and liquidity has a context same as the meaning that decision of construction company’s capital structure supports peaking order theory. However, it was analyzed that it also supports static trade-off theory in part as company’s size or non-debt tax sheild (below average quantile) of a company is also showing a strong influence. Second, as a result of verification whether to changing behavior and non-linearity exist in determinants depending on the general contractor’s over or under borrowed capital structure quantiles, influence of variables showed changing behavior as variables of company’s size, profitability and liquidity of a company have a robust sign in most capital structure quantiles. Non-debt tax sheild variable was analyzed to have a value for practical use as a factor to explain construction company’s capital structure as change of regression coefficient accords with related theory and hypothesis even if non-debt tax sheild variable is capable of securing statistical significance below average borrowing quantile. As a result, this study was able to adopt hypothesis that determinants by general contractor’s capital structure change according to the capital structure quantile. This served as a momentum to exclude the phenomenon that estimated coefficient is concentrated in the estimation method of conditional average of traditional capital structure decision model and identify a new influence which has been overestimated or underestimated so far. Third, as a result of verification whether or not changed behavior of capital structure determinants differs in each sample groups, non-debt tax sheild showed a difference in 0.1 quantile in the verification of the difference of regression coefficient by sample groups. And tangibility variable showed difference between groups in 0.6 quantile. To conclude, this study partially adopted hypothesis that there will exist a difference in capital structure determinants by sample groups. Such a result means that capital structure determinants between sample groups are almost similar. Accordingly, it was analyzed that difference in capital structure decision behavior depending on the difference of company’s size which was suggested in the background of research did not exist within the sample groups. As a result of empirical analysis, this study suggested a variety of implications from the standpoint of management perspective of a company, political perspective of the government and research methodology. From the standpoint of management perspective of a company, it is important in the way that it is helpful to identify decision-making system for decision of construction company’s capital structure by estimating a capital structure decision model of large or small and medium construction companies. In addition, it can be interpreted as a result differentiated from existing researches as the result helps individual construction company to make a decision by alluding to the situation of company’s capital structure by suggesting a spectrum of decision-making system from under to over borrowed capital structure quantiles. From the political perspective of the government, it is considered to be used as preliminary information for analyzing management pattern of large or small and medium construction companies by capital structure quantiles. Most promotion/financial policies are concentrated in small and medium companies suffering from financing, but in case of using capital structure determinants by quantile of small and medium companies, the result of this study is expected to provide help to understand about what environment should be created and for what part should be supported for relaxing financing environment of small and medium companies. Implication produced from the standpoint of research methodology are first, capital structure decision model using traditional estimation method is not capable of understanding capital structure decision making system of a company that uses over, under borrowed capital structure. Thus, this study was able to confirm through a variety of hypothesis verifications that estimation result of quantile regression model can produce more diversified results as compared to the results of estimation by the traditional model. Although it is hard to conclude that quantile regression model is superior, more objective research results are expected to be produced if the two models are properly combined and utilized.; This study judged that repeated bankruptcy of general contractors was caused by their capital structure using excessive borrowed capital under the circumstances of repeated macroeconomic crisis such as IMF foreign exchange crisis in 1997, international financial crisis(2008) and Europe’s financial crisis(2009). Based on the foregoing judgment, this study considered theories and previous researches related to capital structure, and also looked into a research methodology for establishing an empirical model. As a result of consideration of related theories and preceding researches, this study found that researches related to construction industry were very limited and failed in finding a consistent opinion on capital structure determinants. Moreover, it was hard to find a method of empirical analysis in previous researches explaining by what determinants construction company’s over or under borrowing capital structure is decided. Preceding researches were giving priority to substantiation of capital structure determinants and investigation of differences in determinants between sample groups, and researches on changed behavior and non-linearity of capital structure determinants using a quantile regression model are being carried out in other domestic and foreign industries in recent days. Accordingly, this study conducted an empirical analysis on the capital structure determinants by quantile targeting domestic general contractors that are listed·externally audited from 2000 through 2013. Samples used in the empirical analysis include 35 large construction companies, 196 small and medium construction companies, and this study utilized financial information of total 231 construction companies. Objectives of empirical analysis were to verify whether or not gap and similarity exist between behavior (sign) of general contractor’s capital structure determinants and existing theories, verify if changing behavior and non-linearity exist in determinants depending on general contractor’s over or under capital structure quantile and verify whether or not changed behavior of capital structure determinants shows difference between sample groups. To achieve the above study objectives, this study established research hypotheses respectively, and verified the hypotheses after establishing a capital structure decision model by using pooled-ordinary least squares model, generalized least squares model and quantile regression model. Results of empirical analysis can be summarized as following three. First, as a result of verifying whether or not gap and similarity exist between behavior (definite) of construction company’s capital structure determinants and existing theories, it was proved that general contractor’s capital structure determinants support both static trade-off theory and peaking order theory in part. This means that general contractors show a pattern using debt with a advantage of company’ size, but also shows a pattern that uses debt on a shoestring by utilizing internal capital as circumstances require. In the results of hypothesis verification, determinants that explains general contractor’s capital structure effectively are the variables of company’s scale (+), profitability (-), liquidity (-), and each sample group showed a statistically significant and consistent sign in most capital structure quantiles. Meaning that the value of estimated coefficient is high in variables of profitability and liquidity has a context same as the meaning that decision of construction company’s capital structure supports peaking order theory. However, it was analyzed that it also supports static trade-off theory in part as company’s size or non-debt tax sheild (below average quantile) of a company is also showing a strong influence. Second, as a result of verification whether to changing behavior and non-linearity exist in determinants depending on the general contractor’s over or under borrowed capital structure quantiles, influence of variables showed changing behavior as variables of company’s size, profitability and liquidity of a company have a robust sign in most capital structure quantiles. Non-debt tax sheild variable was analyzed to have a value for practical use as a factor to explain construction company’s capital structure as change of regression coefficient accords with related theory and hypothesis even if non-debt tax sheild variable is capable of securing statistical significance below average borrowing quantile. As a result, this study was able to adopt hypothesis that determinants by general contractor’s capital structure change according to the capital structure quantile. This served as a momentum to exclude the phenomenon that estimated coefficient is concentrated in the estimation method of conditional average of traditional capital structure decision model and identify a new influence which has been overestimated or underestimated so far. Third, as a result of verification whether or not changed behavior of capital structure determinants differs in each sample groups, non-debt tax sheild showed a difference in 0.1 quantile in the verification of the difference of regression coefficient by sample groups. And tangibility variable showed difference between groups in 0.6 quantile. To conclude, this study partially adopted hypothesis that there will exist a difference in capital structure determinants by sample groups. Such a result means that capital structure determinants between sample groups are almost similar. Accordingly, it was analyzed that difference in capital structure decision behavior depending on the difference of company’s size which was suggested in the background of research did not exist within the sample groups. As a result of empirical analysis, this study suggested a variety of implications from the standpoint of management perspective of a company, political perspective of the government and research methodology. From the standpoint of management perspective of a company, it is important in the way that it is helpful to identify decision-making system for decision of construction company’s capital structure by estimating a capital structure decision model of large or small and medium construction companies. In addition, it can be interpreted as a result differentiated from existing researches as the result helps individual construction company to make a decision by alluding to the situation of company’s capital structure by suggesting a spectrum of decision-making system from under to over borrowed capital structure quantiles. From the political perspective of the government, it is considered to be used as preliminary information for analyzing management pattern of large or small and medium construction companies by capital structure quantiles. Most promotion/financial policies are concentrated in small and medium companies suffering from financing, but in case of using capital structure determinants by quantile of small and medium companies, the result of this study is expected to provide help to understand about what environment should be created and for what part should be supported for relaxing financing environment of small and medium companies. Implication produced from the standpoint of research methodology are first, capital structure decision model using traditional estimation method is not capable of understanding capital structure decision making system of a company that uses over, under borrowed capital structure. Thus, this study was able to confirm through a variety of hypothesis verifications that estimation result of quantile regression model can produce more diversified results as compared to the results of estimation by the traditional model. Although it is hard to conclude that quantile regression model is superior, more objective research results are expected to be produced if the two models are properly combined and utilized.