Campisi net value method attribution method

Abstract

The invention discloses a Campisi net value method attribution method. The method comprises the following steps: S1, obtaining a fund market data table T1, a bond index market data table T2 and a fund position data table T3 by adopting a database technology; s2, based on the fund quotation data table T1 and the bond index quotation data table T2, selecting a return rate to carry out ADF inspection, and if the return rate is unstable, carrying out differential processing to obtain stable fund and index return rate data; s3, carrying out VIF inspection based on the stable fund and index return rate data; s4, if the collinearity problem exists, PCA processing is carried out on the index return rate sequence, and VIF checking is carried out; s5, performing linear regression on the basis of the obtained index return rate sequence and the fund to obtain a regression coefficient; and S6, based on the regression coefficient, according to the fund position data table T3, calculating the duration of the fund. And the attribution accuracy and stability are ensured.

Description

Technical Field

[0001] The present invention relates to the field of performance attribution analysis, and in particular to a Campisi net worth attribution method. Background Art

[0002] The Campisi model is a model that analyzes the performance of pure bond funds based on holdings data. By decomposing the total returns of the fund and the benchmark into income, treasury bonds, interest rate spreads, and security selection effects (where the security selection effect of the benchmark total return is always zero), and subtracting the corresponding effects of the two combinations, the source of the fund's total return and excess return can be obtained, thereby achieving a quantitative evaluation of the fund manager's investment ability.

[0003] At present, the Campisi net asset value attribution method is mainly achieved by directly performing a linear regression on the returns of all sub-indexes and the returns of the fund. The regression coefficient is then used as the weight of the initial duration of the sub-index, and the sum is used to obtain the initial duration of the fund.

[0004] The method for calculating the initial duration of a fund directly uses the sub-index yield and the fund yield for linear regression, without considering the collinearity and data stationarity issues, resulting in the following problems:

[0005] Reduced stability: Collinearity will increase the estimated variance of model parameters, thereby reducing the stability of the model;

[0006] Reduced explanatory power: When multiple sub-indices are highly correlated, it is difficult to determine which sub-indicator has a greater impact, resulting in reduced explanatory power;

[0007] Reduced accuracy: When the data is not stable, directly applying the regression model may lead to inaccuracies, resulting in a large gap between the results of the initial duration net asset value method and the results of the position method, and failing to truly reflect the initial duration of the fund. Summary of the invention

[0008] In view of the above problems, the present invention is proposed to provide a Campisi net worth attribution method that overcomes the above problems or at least partially solves the above problems.

[0009] According to one aspect of the present invention, a Campisi net worth attribution method is provided, the attribution method comprising:

[0010] Step S1: using database technology to obtain fund market data table T1, bond index market data table T2 and fund holding data table T3;

[0011] Step S2: Based on the fund market data table T1 and the bond index market data table T2, select the yield to perform ADF test. If it is unstable, perform differential processing to obtain stable fund and index yield data;

[0012] Step S3: Perform VIF test based on stable fund and index yield data;

[0013] Step S4: If there is a collinearity problem, perform PCA processing on the index yield series, and then perform VIF test after processing;

[0014] Step S5: Perform linear regression on the obtained index yield sequence and the fund to obtain the regression coefficient;

[0015] Step S6: Based on the regression coefficient and according to the fund holdings data table T3, calculate the duration of the fund.

[0016] Optionally, the fund market data table T1, important features include: daily net value, yield; the bond index market data table T2, important features include: daily remaining term, yield; the fund holdings data table T3, important features include: fund holdings bond yield to maturity, duration index.

[0017] Optionally, the step S2: based on the fund market data table T1 and the bond index market data table T2, selecting the yield to perform an ADF test, and if it is unstable, performing differential processing to obtain stable fund and index yield data specifically includes:

[0018] The time series analysis method using the unit root test ADF is used to determine whether the returns of funds and bond indices have unit roots;

[0019] A unit root means that the trend over time is persistent in the long run and does not converge to a stable mean;

[0020] If the ADF test sequence is non-stationary, the first-order difference of the corresponding yield sequence is performed to obtain a stable yield sequence of the fund and bond index.

[0021] Optionally, the step S3: performing VIF test based on stable fund and index yield data specifically includes:

[0022] The variance inflation factor (VIF) is a measure of the severity of multicollinearity in a multiple linear regression model;

[0023] It represents the ratio of the variance of the regression coefficient estimate to the variance when the independent variables are assumed to be non-linearly correlated. The closer the VIF value is to 1, the lighter the multicollinearity, and vice versa.

[0024] Perform VIF test on bond index yields to obtain bond index with larger VIF.

[0025] Optionally, the step S4: if there is a collinearity problem, PCA processing is performed on the index yield sequence, and then VIF test is performed after the processing, specifically including:

[0026] Eliminate multicollinearity between index returns and extract the main features of collinear factors through PCA or principal component analysis;

[0027] Through the regression of the main features and collinear index yield, the main components are removed from the index yield value and the residual is retained as the new yield value;

[0028] where Y i is the actual observed value, is the regression prediction value, A VIF test is performed on the new index return series for the residuals to ensure that there are no collinearity issues.

[0029] Optionally, the step S5: performing linear regression on the obtained index yield sequence and the fund to obtain the regression coefficient specifically includes:

[0030] Perform multiple linear regression on the fund yield and the new index yield series, and use the least squares method to find the regression coefficient β of each index, which is used as the weight of each index in calculating the fund duration.

[0031] The formula is as follows

[0032] Y=Xβ+ε.

[0033] Optionally, the step S6: based on the regression coefficient and according to the fund holding data table T3, calculating the duration of the fund specifically includes:

[0034] The duration of the fund is equal to the weighted sum of the duration of the bond index, that is, the duration of the bond index multiplied by the regression coefficient. The formula is as follows:

[0035] where β i is the regression coefficient, D i Index duration

[0036] The present invention provides a Campisi net value attribution method, which includes: step S1: using database technology to obtain a fund market data table T1, a bond index market data table T2 and a fund holding data table T3; step S2: based on the fund market data table T1 and the bond index market data table T2, select the yield to perform an ADF test, if unstable, perform differential processing to obtain stable fund and index yield data; step S3: perform a VIF test based on the stable fund and index yield data; step S4: if there is a collinearity problem, perform PCA processing on the index yield sequence, and then perform a VIF test after processing; step S5: based on the obtained index yield sequence and the fund, perform linear regression to obtain a regression coefficient; step S6: based on the regression coefficient, calculate the duration of the fund according to the fund holding data table T3. Solve the data stationarity and collinearity problems, obtain the regression coefficient by the linear regression method, and calculate the initial duration of the fund by the regression coefficient to ensure the accuracy and stability of the attribution.

[0037] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0039] Figure 1 A flowchart of a Campisi net worth attribution method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0041] The terms "comprises" and "having" and any variations thereof in the description embodiments, claims and drawings of the present invention are intended to cover non-exclusive inclusions, for example, including a series of steps or units.

[0042] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0043] Example 1

[0044] An improved Campisi net worth attribution method based on ADF and PCA includes:

[0045] Step S1: Obtain the fund market data table T1 through database technology, where important features include daily net value and yield, the bond index market data table T2, where important features include daily remaining term and yield, and the fund holding data table T3, where important features include indicators such as yield to maturity and duration of fund holding bonds;

[0046] Step S2: Based on step S1, tables T1 and T2 are obtained, and the yield is selected for ADF test. If it is unstable, differential processing is performed to obtain stable data;

[0047] The time series analysis method using the unit root test (ADF) is used to determine whether the returns of the fund and bond index have unit roots (i.e., non-stationarity). A unit root means that the trend over time is persistent in the long term and does not converge to a stable mean. If the ADF test sequence is non-stationary, the first-order difference of the corresponding return sequence is performed to obtain a stable return sequence of the fund and bond index.

[0048] Step S3: Perform VIF test based on the stable fund and index yield data obtained in step S2;

[0049] Variance inflation factor (VIF) is a measure of the severity of multicollinearity in a multivariate linear regression model. It represents the ratio of the variance of the regression coefficient estimate to the variance when the independent variables are assumed to be nonlinearly correlated. The closer the VIF value is to 1, the lighter the multicollinearity, and vice versa. Perform a VIF test on the bond index yield to find the bond index with a larger VIF.

[0050] Step S4: If there is a collinearity problem based on step S3, PCA processing is performed on the index yield series, and then VIF test is performed after processing;

[0051] To eliminate multicollinearity between index returns, the main features of collinear factors are extracted through PCA (principal component analysis), and then the main features are regressed with the collinear index returns. The main components are removed from the index return value and the residual is retained as the new return value.

[0052] where Y i is the actual observed value, is the regression prediction value, A VIF test is performed on the new index return series for the residuals to ensure that there are no collinearity issues.

[0053] Step S5: Perform linear regression on the index yield series obtained in step S4 and the fund to obtain the regression coefficient;

[0054] The fund yield is subjected to multiple linear regression with the new index yield sequence, and the regression coefficient β of each index is obtained by the least squares method, which is used as the weight of each index in calculating the fund duration. The formula is as follows

[0055] y=β 0 +β 1 x 1 +β 2 x 2 +β 3 x 3 +ε

[0056] Y=Xβ+ε

[0057] Step S6: Based on the regression coefficient obtained in step S5, calculate the duration of the fund according to the duration data in T3.

[0058] The duration of the fund is equal to the weighted sum of the duration of the bond index, that is, the duration of the bond index multiplied by the regression coefficient. The formula is as follows:

[0059] where β i is the regression coefficient, D i Index duration

[0060] Example 2

[0061] An improved Campisi net worth attribution method based on ADF and PCA, including:

[0062] Based on the database data, we get the fund market data table T1 and the bond index market data table T2 as follows

[0063] Fund Code date Adjusted net value Yield 000033.OF 20230704 1.5175 0.013181 000033.OF 20230705 1.5184 0.059308 ... ... ... ...

[0064] date Index Code Remaining term Yield 2023-07-03 CBA00111.CB 0.4587 0.034209 2023-07-04 CBA00111.CB 0.4567 0.008615 ... ... ... ...

[0065] In step S2, an ADF test is performed. If it is unstable, differential processing is performed to obtain stable data.

[0066] ADF test results:

[0067]

[0068]

[0069] The result after data differentiation:

[0070] date CBA02711.CB CBA02721.CB CBA02731.CB CBA02741.CB CBA02751.CB 20230705 -0.006570 0.015812 0.029425 0.023708 -0.011365 20230706 0.017221 0.029969 0.037398 0.034875 0.034098 ... ... ... ... ... ...

[0071] Step S3 performs VIF test on the obtained stable fund and index yield data:

[0072] Index Code VIF CBA02711.CB 1.264462 CBA02721.CB 19.882192 CBA02731.CB 28.642487

[0073] Step S4 performs PCA processing on the index yield series with collinearity:

[0074] date CBA02711.CB CBA02721.CB CBA02731.CB CBA02741.CB CBA02751.CB 20230705 -0.007279 -0.006570 -0.011365 0.013541 0.026344 20230706 0.038625 0.017221 0.034098 0.019634 0.023378 ... ... ... ... ... ...

[0075] Step S5 obtains the regression coefficient through linear regression and in step S6 calculates the fund duration.

[0076] Beneficial effects: The present invention will be based on fund holdings, market and basic data as well as market data of bond indexes, and will perform data stationarity test through ADF test and feature extraction through PCA principal component analysis to solve data stationarity and collinearity problems, and obtain regression coefficients through linear regression method. The initial duration of the fund is calculated through the regression coefficients to ensure the accuracy and stability of attribution.

[0077] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A Campisi net worth attribution method, characterized in that: The attribution methods include: Step S1: using database technology to obtain fund market data table T1, bond index market data table T2 and fund holding data table T3; Step S2: Based on the fund market data table T1 and the bond index market data table T2, select the yield to perform ADF test. If it is unstable, perform differential processing to obtain stable fund and index yield data; Step S3: Perform VIF test based on stable fund and index yield data; Step S4: If there is a collinearity problem, perform PCA processing on the index yield series, and then perform VIF test after processing; Step S5: Perform linear regression on the obtained index yield sequence and the fund to obtain the regression coefficient; Step S6: Based on the regression coefficient and according to the fund holdings data table T3, calculate the duration of the fund.

2. A Campisi net worth attribution method according to claim 1, characterized in that: The fund market data table T1, important features include: daily net value, yield; the bond index market data table T2, important features include: daily remaining term, yield; the fund holdings data table T3, important features include: fund holdings bond maturity yield, duration index.

3. A Campisi net worth attribution method according to claim 1, characterized in that: The step S2: based on the fund market data table T1 and the bond index market data table T2, selecting the yield to perform ADF test, if it is unstable, performing differential processing to obtain stable fund and index yield data specifically includes: The time series analysis method using the unit root test ADF is used to determine whether the returns of funds and bond indices have unit roots; A unit root means that the trend over time is persistent in the long run and does not converge to a stable mean; If the ADF test sequence is non-stationary, the first-order difference of the corresponding yield sequence is performed to obtain a stable yield sequence of the fund and bond index.

4. A Campisi net worth attribution method according to claim 1, characterized in that: The step S3: performing VIF testing based on stable fund and index yield data specifically includes: The variance inflation factor (VIF) is a measure of the severity of multicollinearity in a multiple linear regression model; It represents the ratio of the variance of the regression coefficient estimate to the variance when the independent variables are assumed to be non-linearly correlated. The closer the VIF value is to 1, the lighter the multicollinearity, and vice versa. Perform VIF test on bond index yields to obtain bond index with larger VIF.

5. A Campisi net worth attribution method according to claim 1, characterized in that: The step S4: if there is a collinearity problem, PCA processing is performed on the index yield sequence, and then VIF test is performed after the processing, which specifically includes: Eliminate multicollinearity between index returns and extract the main features of collinear factors through PCA or principal component analysis; Through the regression of the main features and collinear index yield, the main components are removed from the index yield value and the residual is retained as the new yield value; where Y i is the actual observed value, is the regression prediction value, A VIF test is performed on the new index return series for the residuals to ensure that there are no collinearity issues.

6. A Campisi net worth attribution method according to claim 1, characterized in that: The step S5: performing linear regression on the obtained index yield sequence and the fund to obtain the regression coefficient specifically includes: Perform multiple linear regression on the fund yield and the new index yield series, and use the least squares method to find the regression coefficient β of each index, which is used as the weight of each index in calculating the fund duration. The formula is as follows Y=Xβ+ε 7. A Campisi net worth attribution method according to claim 1, characterized in that: The step S6: based on the regression coefficient and according to the fund holding data table T3, calculating the duration of the fund specifically includes: The duration of the fund is equal to the weighted sum of the duration of the bond index, that is, the duration of the bond index multiplied by the regression coefficient. The formula is as follows: where β i is the regression coefficient, D i is the index duration.

Images