The Establishment of Risk Evaluation Index System for Small- and Medium-Sized Agency Bookkeeping Companies

In recent years, China's economy has developed rapidly; many small companies have risen rapidly; and the tax system has become more and more standardized. Because many small businesses cannot afford to hire full-time accountants, they opt to outsource accounting services, giving small- and medium-sized bookkeeping firms a large market space. However, these opportunities also bring huge operational risks to small- and medium-sized bookkeeping companies. The purpose of this research is to help such enterprises carry out risk management and reduce operational risks. This study uses an analytic hierarchy process and a fuzzy comprehensive assessment approach to successfully combine quantitative and qualitative analysis and create a multilevel analysis structure model of the risk management evaluation index system of small- and medium-sized agency accounting firms. The structural model is verified by a case, the specific risk score of each factor is calculated through the scores of 20 experts, and the importance of risk is judged according to the size of the score, indicating that the structural model is feasible.


Introduction
In order to promote the healthy development of the agency bookkeeping industry, the "Administrative Measures for Agency Bookkeeping" was officially implemented in May 2016, which once again standardized the institutional management of the agency bookkeeping industry [1], but this did not fundamentally solve the problem of industrial development.
e Financial Accounting (2018) No. 32 document particularly emphasized the need to strictly manage agency bookkeeping institutions, establish and improve the integrity incentive and untrustworthy punishment mechanism for the agency bookkeeping industry, and regulate the management of industry associations [2]. In March 2019, the Ministry of Finance issued relevant regulations to simplify the application materials for agency bookkeeping qualifications, shorten the statutory approval time limit, and stimulate the vitality of market entities [3].
e bookkeeping industry has experienced nearly 30 years of development, from the traditional manual mode to the computerized mode [4]. Due to the development of artificial intelligence, the accounting and taxation of agency bookkeeping companies have gradually become intelligent [5], which has brought greater influence to small-and mediumsized bookkeeping companies, and the problem of risk management has become increasingly prominent [6].

Basic Method eory.
In the "Comprehensive Risk Management Framework" released by COSO in 2003, comprehensive risk management includes three dimensions [7] as shown in Figure 1.
e basic process of comprehensive risk management is divided into five steps. e implementation of risk management is inseparable from the communication of information [8].
erefore, a complete risk management information system must be established [9], as shown in Figure 2.
e formation process of agency bookkeeping risk: risk event-risk taker-risk loss [10], as shown in Figure 3.

Establishment of the Risk Evaluation Index
System for Small-and Medium-Sized Agency Bookkeeping Companies

Establishment of the Risk Assessment Index System.
In practical applications, there are many types of risk factor identification methods [19], mainly including the following: (1) brainstorming method, (2) analysis process method [20], (3) analysis of relevant scenarios, (4) risk decomposition method [21], and (5) editing event tree method [22]. According to the actual situation of small-and mediumsized agency bookkeeping companies, this paper adopts the brainstorming method to identify the risk factors of these companies.
Organized and sent 20 experts to conduct in-depth research on small-and medium-sized bookkeeping companies. rough the discussion, summarize the main risks in five aspects: policy and legal risk, industry competition risk, information technology risk, undertaking business risk, and practitioner risk [23]. In the case of ensuring the comprehensiveness of the risk evaluation system, the risk evaluation index system in Table 1 is summarized and determined [24].

Establishment of Risk Assessment Set.
In order to make the evaluation effect clearer, the evaluation using gradients within a range of the evaluated risk factors is usually adopted [25]. Five different continuous grade categories are selected for the agency bookkeeping risk evaluation set, and the five evaluation results are "no risk," "small risk," "average risk," "high risk," and "huge risk," Specifically, it is represented by V � {V1, V2, V3, V4, V5}, in which the five evaluation results are, respectively, corresponding to the scores of "20," "40," "60," "80," and "100." Higher scores indicate greater risk, as shown in Table 2.

Establishing the Judgment Matrix.
Use an appropriate scale to construct a judgment matrix by comparison [26] and establish a comparative judgment matrix A for the risk indicators after statistical analysis: (1)

Calculation of Criterion Layer
Weights. e calculation of the weight of the criterion layer can be divided into three steps: one is to normalize the risk judgment matrix A of the criterion layer of the company to obtain a new matrix [27]. e weights are obtained by normalization [28]. e third is the matrix consistency test [29].
(1) Normalize each column of the A matrix to obtain a new matrix Z: (2) e eigenvectors are obtained by summing each row of the matrix, and then the weights are obtained by normalizing the eigenvectors: WA � (WA 1, WA 2, WA 3, WA 4, WA 5). From this, it can be concluded that WA1 to WA5 are the weight ratios of each risk factor index in each criterion layer, and the next work is to take a consistency check on the weight of each index obtained.

(3) Matrix consistency test
We calculate the largest eigenroot as follows: where λmax represents the largest eigenvalue, A represents the corresponding matrix, W represents the eigenvector, and i represents the corresponding element in the matrix. e CI scale is determined based on the characteristic root. e formula is as follows: We calculate the consistency index of the judgment matrix as follows: Factor CR agreement was calculated according to the RI corresponding to the CI. In fact, RI is a constant, which represents the average random consistency index. According to the order of the matrix, the ratio of the matrix consistency index CI and the average random consistency index RI of the same order can be queried in the table, which is called the random consistency ratio. If the condition of CR < 0.10 is satisfied, it can indicate that the judgment matrix has passed the consistency check, and if it is not satisfied, it means that the consistency check has not passed [30].
We calculate the random consistency ratio as follows:

Calculation of Index Layer Weights and Comprehensive
Weights. When determining other weights, you can also collect and analyze data and use the same method as above to calculate the standard layer judgment matrix weight. According to the formula: the comprehensive weight of the indicator layer � the weight of the indicator layer × the weight of the criterion layer, the comprehensive weight of each indicator of the company's indicator layer can be obtained by calculation.

Application of the Fuzzy Comprehensive Evaluation
Method. After the weight of each risk factor index is obtained, the membership degree of the fuzzy comprehensive evaluation set should be calculated next, and the fuzzy comprehensive evaluation should be implemented by combining the weight and the membership degree [31]. e determination of membership degree is highly subjective, requiring each evaluation expert to be careful [32]. e membership matrix function usually refers to a new matrix that is synthesized after the evaluation indicators of all matrices are rated by membership [33], as shown in the following formula: R can be regarded as a mapping of the matrix membership evaluation set, and the membership degree is treated as a condition of fuzzy operation, and the fuzzy evaluation result can be obtained by multiplying it by the weight. e specific calculation formula is shown in the following formula: Formula (8) can be used to calculate the fuzzy comprehensive evaluation score of the five matrices of the criterion layer "A1" to "A5." From the above calculation ideas, usually use "very important, relatively important, generally important, less important, very unimportant," the 5-level satisfaction evaluation level that makes a reasonable   evaluation of all evaluation indicators, so as to obtain the membership degree. Twenty experts selected Appendix B and obtained the evaluation frequency table after statistics. e corresponding membership degree matrix can be obtained by obtaining the membership degree of each index.

Case Validation
4.1. Case Situation. HIG bookkeeping company was established in 2013. It is a typical small-and medium-sized bookkeeping company in China. e company has a registered capital of 1 million yuan. Its business scope includes accounting business consulting services, accounting consultants, agency bookkeeping services [34], agency financial and tax reporting services, agent for industrial and commercial registration, enterprise annual report service, agent for various licenses, and agent for trademark registration and patent application [35].
After the discussion of 20 experts, the final conclusion is drawn to the HIG company's agency bookkeeping risk formation table, as shown in Table 3.

Constructing the Judgment Matrix.
A total of 20 experts were invited for this case, including 3 managers and deputy managers of the company; one person in charge of each department of Administration Department, Operation Department, Business Department, and Finance Department; and 13 external financial experts. After statistical analysis, a comparison (judgment) matrix A is established for the risk indicators:

Calculation of Criterion Layer Weights
(1) Normalize each column of the A matrix to obtain a new matrix Z as follows: (2) e eigenvectors are obtained by summing each row of the matrix, and then the weights are obtained by normalizing the eigenvectors [36]: We calculate the largest eigenroot as follows: Here, λmax � 5.2541. We calculate the consistency index of the judgment matrix as follows: Here, CI � 0.0635. e average random consistency index is shown in Table 4. We calculate the random consistency ratio as follows: H CR � 0.0567 < 0.10. erefore, it can be determined that WA � (0.1827, 0.0596, 0.2522, 0.0977, 0.4078) is the criterion layer weight that satisfies the consistency test conditions.

Calculation of Indicator Layer
Weights. Using the same method as above, the standard layer judgment matrix weights are calculated as shown in Tables 5-9, respectively. λmax � 3.0735 and CR � 0.0707 < 0.10. Indicator layer weights for policy legal risks: WA1 � (0.1717, 0.4414, 0.3869).

Calculation of Comprehensive Weight of Index Layer.
According to the formula: the comprehensive weight of the indicator layer � the weight of the indicator layer × the weight of the criterion layer, the comprehensive weight of each indicator in the indicator layer of HIG can be obtained by calculation, as shown in Table 10.
It can be found from Table 10 that the top three comprehensive weights of the indicator layer are professional skills risk, financial software technology risk, and professional ethics risk.  Table 11. Taking the membership evaluation of a11 as an example, it is very important for 2 experts to choose a11; the choice of 8 experts is more important; the choice of 8 experts is generally important; the choice of 2 experts is not very important; and the choice of no experts is very unimportant.

Application of the Fuzzy Comprehensive Evaluation
en the membership degree of a11 is as follows: r11 � (0.1, 0.4, 0.4, 0.1, 0), and so on, the corresponding membership degree matrix can be obtained by obtaining the membership degree of each index.

Fuzzy Operations.
Taking the criterion-level indicator of policy and legal risk as an example, according to formula   Computational Intelligence and Neuroscience Equally important e M metric is as important as the N metric 2 Importance is between 1 and 3 3 Slightly important e M index is slightly more important than the N index 4 e importance is between 3 and 5 5 Obviously important e M index is significantly more important than the N index 6 e importance is between 5 and 7 7 Much more important M index is much more important than the N index 8 e importance level is between 7 and 9 9 Extremely important M index is extremely important than N index Note: if the ratio of the M index to the N index is a, then the ratio of the N index to the M index is 1/a.      on the top three risk factors ranked by the comprehensive weight of the indicator layer, and focus on the response.

Discussion
e main research methods of this study are the analytic hierarchy process and the fuzzy comprehensive evaluation method. Analytic hierarchy process is a decision analysis method that combines qualitative and quantitative analysis to solve complex multiobjective problems; the relative importance of each decision-making scheme is given; the weight of each standard of each decision-making scheme is reasonably given; and the weights are used to obtain the superior and inferior order of each scheme, which can be effectively applied to those problems that are difficult to solve by quantitative methods. e fuzzy comprehensive evaluation method is a comprehensive evaluation method that transforms qualitative evaluation into quantitative evaluation through the membership degree theory of fuzzy mathematics. e above two methods are greatly influenced by personal subjectivity, have certain limitations, and need to be improved.

Conclusions
ere are now few small-and medium-sized bookkeeping firms, and the industry's future prospects are unclear. Future social and economic development are inextricably linked to long-term sustainable development. As a result, there are fewer studies on these organisations that are relevant, and risk management studies are more beneficial.
is study uses an analytic hierarchy process and a fuzzy comprehensive assessment approach to successfully combine quantitative and qualitative analysis and create a multilevel analysis structure model of the risk management evaluation index system of small-and medium-sized agency accounting firms. e specific risk score of each factor is calculated through the scores of 20 experts, and the importance of risk is judged according to the size of the score, which helps the company formulate corresponding risk measures. is research provides a reference for the risk management of enterprises in this industry and has certain research value.