XBRL财务报告分类标准：微观结构、质量评价和改进方案

发布时间：2018-09-13 13:57

【摘要】：自美国注册会计师查尔斯霍夫曼1998年开创性的将有丰富语义表达能力的可扩展标记语言技术应用于财务报告，并逐渐形成了可扩展商业报告语言（eXtensible Business Reporting Language，，简称XBRL）的概念以来，XBRL在全球范围内实践和发展经历了十五载。目前，XBRL财务报告分类标准（简称分类标准）的发展趋势已经由制定和实施逐步过渡到评价和改进。制定和实施方面现有的研究已经硕果累累，但是关于评价和改进的研究还有不少问题一直有待完善。第一个问题是“分类标准的微观结构是什么？”，它是评价和改进研究的基础。现有的研究者认为分类标准的最基本单元是财务信息元素，但是随着维度建模方法的大量使用，分类标准的微观结构正在发生着变化，该问题的解决不仅有助于深化了财务信息元素理论，而且为评价分类标准的质量指明了研究方向。第二个问题是“如何评价分类标准的质量？”，它是评价和改进研究的核心。现有的研究者多是从分类标准的完整性角度来评价分类标准的质量，而缺乏从创建和扩展视角评价分类标准质量。对创建和扩展分类标准的不同模式进行评价的目的是发现它们的优点和缺点，为进一步完善分类标准提供定性和定量的依据。第三个问题是“如何改进分类标准的质量？”，它是评价和改进研究的成果。现有的研究者就通用层级分类标准的改进提出过不少政策建议和实施框架，但是无论怎么改进，都不可避免地要在分类标准的完整性和可比性之间做出取舍，这种巴别塔式的困境随着行业层级分类标准的引入有所改观，但现有研究尚缺乏创建行业层级分类标准的有效方法，这为分类标准的质量改进蒙上了一层阴影。一套可操作的创建行业层级分类标准的方法将有助于改进分类标准的质量，改善财务和其他经济信息在经济实体之间的相互交换，以满足报告使用者对高质量财务报告的要求。这三个基本问题是实践中分类标准制定者、实例文档创建者、财务信息监管者乃至投资者都关注的问题，同时也是XBRL财务报告理论研究的焦点。本文以微观经济、财务会计，以及集合论、矩阵论、概率论和数理统计等理论为基础，结合实证检验的方法，围绕评价与改进分类标准质量为主题展开研究。第一个研究问题是“分类标准的微观结构是什么？”。采用逆向工程的研究方法，对分类标准中信息元素的遴选和归集方法做了对比分析。从创建和扩展的视角分析和比较了分类标准的微观结构，对分类标准的微观结构进行了形式化描述。第二个研究问题是“如何评价分类标准的质量？”。从创建和扩展分类标准的角度，分别构造了衡量分类标准质量的评价标准和指标体系。关于分类标准的创建质量，构建了创建质量测度，针对元组模式和维度模式，深化了该创建质量模型，最后以财务报表附注的信息元素为样本，对不同的创建模式（元组模式选取上交所上市公司分类标准，维度模式选取通用分类标准）进行了创建效率、语义信息完整性和创建质量的度量和评价。关于分类标准的扩展质量，构建了扩展质量测度，针对直接扩展模式和行业扩展模式，深化了该扩展质量模型，最后以34家石油行业上市公司财务报告附注的信息元素为样本，对不同的扩展模式进行了完整性、效率性和可比性的度量、评价和稳健性检验。第三个研究问题是“如何改进分类标准的质量？”。从实务角度提出了一套可操作的创建行业层级分类标准的理论和方法，并以资本市场中占比最高的制造业为例，选取了153家制造业上市公司财务报告附注信息做为样本，来创建该行业分类标准。本文的主要研究结论如下： 1、提出了在不同的创建模式下分类标准的最基本单元不同的观点。元组模式下财务信息元素是构建分类标准的最基本单元；维度模式下结构信息元素（表头、轴成员和列报项目等）是构建分类标准的最基本单元；由轴成员和列报项目信息元素构造了影子财务信息元素。 2、从创建质量方面看，分类标准的维度模式优于元组模式；从扩展质量方面看，分类标准的行业扩展模式优于直接扩展模式。创建质量方面，在假定信息元素空间的单位创建成本相同，信息元素空间的单位披露收益相同，创建效率和语义信息完整性同等重要的前提下，总体上维度模式分类标准的创建质量优于元组模式分类标准；维度模式分类标准的创建效率优于元组模式分类标准；元组模式分类标准的语义信息完整性优于维度模式分类标准。扩展质量方面，不同扩展模式的完整性、效率性和可比性在统计上有显著的差异；行业扩展模式在经济上有显著的完整性、效率性和可比性优势。平均而言，行业扩展模式的可比性比直接扩展模式好四成左右，而且优势不会随着信息元素集合的缩小而发生变化，反映出评价上市公司可比性的测度模型是稳健的。 3、提出了一套创建行业分类标准的方法，创建了制造业分类标准。在拓展信息元素空间理论的基础上，以报告使用者决策有用性为目标，提出了基于频数遴选信息元素的方法，构造了经济意义上确定频数的可比性效用最优理论模型和统计意义确定频数的直观方法。以制造业样本为例，分别计算了整体信息元素的扩展频数、扩展密度、查询成本调节因子、平均可比性、修正可比性、累计扩展元素数量和累计扩展元素比例等相关指标，通过可比性效用最优理论确定了最优扩展频数为66；最后选取扩展频数大于等于66的信息元素集合，创建了制造业行业分类标准。本文的创新点主要体现在以下三个方面： 1、拓展了分类标准财务信息元素理论，重构了信息元素空间理论。引入集合论描述了分类标准的微观结构，比较了分类标准的创建模式和扩展模式，拓展了现有的财务信息元素理论，为评价分类标准的质量奠定了理论基础。将信息元素的频数引入到信息元素空间中，重新定义了信息元素空间，提出了频数-密度空间和频数-概率密度空间。将元素空间理论由一维扩展到多维，将元素域扩展到了元素-频数域、频数-密度域和频数-概率密度域；建立了元素-频数-密度-概率密度的函数映射关系；为从实务角度遴选信息元素，构造行业分类标准奠定了理论基础。 2、构造了度量分类标准的创建和扩展质量测度，评价了分类标准的创建和扩展质量。创建质量方面，基于成本收益原则和信息完整性构建了创建质量测度，尝试从创建效率、语义信息完整性和整体创建质量三个角度评价了分类标准的不同创建模式，为创建分类标准的实践提供量化依据。扩展质量方面，与目前研究者们基本上采用信息匹配的方法来评价分类标准的完整性不同，本文引入频数统计法评价分类标准的扩展质量；采用累计扩展量来评价分类标准的完整性；采用累计复用量来评价分类标准的效率性；以累计复用量为基础构造了分类标准的可比性测度；在假设分类标准扩展质量中完整性、效率性和可比性同等重要的前提下，构造了整体扩展质量测度；尝试从完整性、效率性和可比性三个角度评价了分类标准的不同扩展模式，为扩展分类标准的实践提供了量化依据。 3、构造了统计意义上依指定概率和经济意义上依可比性效用最优筛选信息元素、创建行业分类标准的方法。从实务披露角度出发，采用频数法遴选分类标准的信息元素，构造行业分类标准。对财务报告附注中的信息元素进行频数统计；通过转换得到对应频数的复用密度和扩展密度，通过单位化变换得到对应频数的概率密度分布；按一定的统计意义或经济意义确定频数下限，筛选出大于等于该频数下限的信息元素构成行业分类标准。其中，统计意义通过信息元素的比例直接筛选信息元素，经济意义通过效用最优来间接筛选信息元素。该方法弥补了实务法遴选信息元素方法上的不足。
[Abstract]:Since Charles Hoffman, an American CPA, pioneered the application of extensible markup language (XBRL) in financial reporting in 1998, and gradually formed the concept of eXtensible Business Reporting Language (XBRL), XBRL has been practiced and developed worldwide. For fifteen years.
At present, the development trend of XBRL Financial Reporting Classification Standard (XBRL Financial Reporting Classification Standard) has gradually changed from formulation and implementation to evaluation and improvement. Current researchers believe that the basic unit of classification criteria is financial information elements, but with the extensive use of dimension modeling methods, the microstructure of classification criteria is changing, and the solution of this problem not only helps deepen the theory of financial information elements, but also helps to deepen the theory of financial information elements. The second question is "how to evaluate the quality of classification criteria?", which is the core of evaluating and improving research. Most of the existing researchers evaluate the quality of classification criteria from the perspective of the integrity of classification criteria, but lack of evaluating the quality of classification criteria from the perspective of creation and expansion. The purpose of creating and expanding different models of classification standards is to discover their strengths and weaknesses and to provide qualitative and quantitative basis for further improvement of classification standards. Many policy suggestions and implementation frameworks have been put forward for improvement, but no matter how it is improved, it is inevitable to make a choice between the integrity and comparability of classification standards. This kind of Babel-style dilemma has been improved with the introduction of industry-level classification standards, but the existing research still lacks the effectiveness of establishing industry-level classification standards. A set of operational methods for creating industry-level classification standards will help to improve the quality of classification standards and improve the exchange of financial and other economic information between economic entities to meet the reporting users'requirements for high-quality financial reporting. This problem is the focus of XBRL financial reporting theory research, which is concerned by the classification standard setters, case document creators, financial information supervisors and even investors in practice.
Based on the theories of micro-economy, financial accounting, set theory, matrix theory, probability theory and mathematical statistics, and combined with the method of empirical test, this paper focuses on the evaluation and improvement of the quality of classification standards.
The first research question is "what is the micro-structure of classification criteria?". The method of selecting and collecting information elements in classification criteria is compared and analyzed by means of reverse engineering. The micro-structure of classification criteria is analyzed and compared from the perspective of creation and expansion, and the micro-structure of classification criteria is formally described. Say.
The second research question is "how to evaluate the quality of classification criteria?". From the angle of creating and expanding classification criteria, the evaluation criteria and index systems for evaluating the quality of classification criteria are constructed respectively. Finally, taking the information elements of financial statements as samples, we measure and evaluate the creation efficiency, semantic information integrity and creation quality of different creation modes (tuple mode selecting the classification standards of listed companies on the Shanghai Stock Exchange, dimension mode selecting the general classification standards). Finally, we take the information elements of 34 listed companies in the oil industry as samples to measure the completeness, efficiency and comparability of different expansion modes, evaluate and test their robustness.
The third research question is "how to improve the quality of classification standards?". From the practical point of view, this paper puts forward a set of operable theory and method of establishing industry hierarchical classification standards, and takes the manufacturing industry with the highest proportion in the capital market as an example, selects the financial report annotations of 153 manufacturing listed companies as samples to create the industry. Classification standard.
The main conclusions of this paper are as follows:
1. Put forward different viewpoints on the basic unit of classification standard under different creation modes. Financial information element is the basic unit to construct classification standard under tuple mode; structural information element (table head, axis member and presentation item, etc.) under dimension mode is the basic unit to construct classification standard; axis member and presentation item are the basic unit to construct classification standard. The information element constructs the shadow financial information element.
2. In terms of creating quality, the dimension model of classification standard is superior to tuple model; in terms of extending quality, the industry expansion model of classification standard is superior to direct expansion model. On the premise that information integrity is equally important, the quality of creating dimension schema classification standard is better than tuple schema classification standard on the whole; the efficiency of creating dimension schema classification standard is better than tuple schema classification standard; the semantic information integrity of tuple schema classification standard is better than dimension schema classification standard. There are significant statistical differences in the completeness, efficiency and comparability of the extended model; there are significant economic completeness, efficiency and comparability advantages of the industrial expansion model. The change shows that the measurement model for evaluating the comparability of listed companies is robust.
3, we put forward a set of methods to establish industry classification standard and set up a classification standard for manufacturing industry.
On the basis of expanding the theory of information element space and aiming at reporting the usefulness of users in decision-making, a method of selecting information elements based on frequency is proposed. The optimal theoretical model of comparability utility for determining frequency in economic sense and the intuitive method for determining frequency in statistical sense are constructed. Taking manufacturing sample as an example, the whole system is calculated separately. Based on the theory of comparability utility optimization, the optimal spread frequency is determined to be 66. Finally, the set of information elements whose spread frequency is greater than or equal to 66 is selected and created. The classification standard of manufacturing industry.
The innovation of this paper is mainly reflected in the following three aspects:
1. Extended the theory of financial information elements of classification standards and reconstructed the theory of information element space. Introduced set theory to describe the micro-structure of classification standards, compared the creation mode and expansion mode of classification standards, expanded the existing theory of financial information elements, laid a theoretical foundation for evaluating the quality of classification standards. Frequency is introduced into information element space, and the information element space is redefined. Frequency-density space and frequency-probability density space are proposed. Element space theory is extended from one dimension to multi-dimension, element domain to element-frequency domain, frequency-density domain and frequency-probability density domain. The function mapping relation of degree has laid a theoretical foundation for selecting information elements from practical point of view and constructing industry classification standards.
2. Constructed the creation and extension quality measures of the measurement classification standards, evaluated the creation and extension quality of the classification standards. Extended quality is different from the method of information matching that researchers basically use to evaluate the integrity of classification standards. This paper introduces frequency statistics to evaluate the extended quality of classification standards, and uses cumulative expansion to evaluate the integrity of classification standards. The efficiency of classification criteria is evaluated by cumulative reuse quantity; the comparability measure of classification criteria is constructed on the basis of cumulative reuse quantity; the whole extended quality measure is constructed on the premise that integrity, efficiency and comparability are equally important in the extended quality of classification criteria; and the integrity, efficiency and comparability are attempted This paper evaluates the different extension modes of classification standards from three perspectives, which provides a quantitative basis for the practice of extending classification standards.
3. Construct the method of selecting the information elements according to the specified probability and the comparability utility in the economic sense to create the industry classification standard. The multiplexing density and spreading density of corresponding frequencies are obtained by conversion, and the probability density distribution of corresponding frequencies is obtained by unit transformation; the lower limit of frequencies is determined according to certain statistical meaning or economic meaning, and the information elements larger than or equal to the lower limit of the frequency are selected to form the industry classification standard. This method makes up for the shortcomings of practical method in selecting information elements.
【学位授予单位】：上海交通大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：F232

【参考文献】