Các nhân tố ảnh hưởng đến hiệu quả hoạt động của ngân hàng thương mại cổ phần của việt nam khóa luận đại học chuyên ngành tài chính ngân hàng lê đức duy võ thị ngọc hà người hướng dẫn khoa học

NGÂN HÀNG NHÀ NƢỚC VIỆT NAM BỘ GIÁO DỤC VÀ ĐÀO TẠO TRƢỜNG ĐẠI HỌC NGÂN HÀNG TP HỒ CHÍ MINH - KHÓA LUẬN TỐT NGHIỆP CÁC NHÂN TỐ ẢNH HƢỞNG ĐẾN HIỆU QUẢ HOẠT ĐỘNG CỦA NGÂN HÀNG THƢƠNG MẠI CỔ PHẨN TẠI VIỆT NAM Mã số: 7340201 Sinh viên thực hiện: LÊ ĐỨC DUY Lớp: HQ4_GE04 Khóa học: CLC04 MSSV: 030632160435 Giảng viên hƣớng dẫn: ThS Võ Thị Ngọc Hà Tp Hồ Chí Minh, tháng - 2020 NGÂN HÀNG NHÀ NƢỚC VIỆT NAM BỘ GIÁO DỤC VÀ ĐÀO TẠO TRƢỜNG ĐẠI HỌC NGÂN HÀNG TP HỒ CHÍ MINH - ĐỀ CƢƠNG KHÓA LUẬN CÁC NHÂN TỐ ẢNH HƢỞNG ĐẾN HIỆU QUẢ HOẠT ĐỘNG CỦA NGÂN HÀNG THƢƠNG MẠI CỔ PHẨN TẠI VIỆT NAM Mã số: 7340201 Sinh viên thực hiện: LÊ ĐỨC DUY Lớp: HQ4_GE04 Khóa học: CLC04 MSSV: 030632160435 Giảng viên hƣớng dẫn: ThS Võ Thị Ngọc Hà Tp Hồ Chí Minh, tháng - 2020 TĨM TẮT Khóa luận phân tích nhân tố ảnh hƣởng đến hiệu hoạt động 24 ngân hàng thƣơng mại thơng qua báo cáo tài đƣợc cơng bố khoảng thời gian từ 2007-2019 Nghiên cứu lựa chọn tỷ lệ sinh lời tài sản (ROA) tỷ lệ sinh lời vốn chủ sở hữu (ROE) làm biến phụ thuộc Thơng qua kết mơ hình hồi quy đa biến mơ hình tác động cố định (FEM) đƣợc kết luận phù hợp Các biến độc lập mơ hình gồm (1) Quy mơ tài sản, (2) Tỷ lệ vốn chủ sở hữu, (3) Tỷ lệ dự phòng rủi ro, (4) Tỷ lệ tổng chi phí hoạt động, (5) Đa dạng hóa thu nhập, (6) Tăng trƣởng tín dụng, (7) Tăng trƣởng GDP, (8) CPI, (9) biến giả Crisis Kết kiểm định cho thấy khơng có tƣợng tự tƣơng quan nhƣng tồn vấn đề phƣơng sai thay đổi dù thực biện pháp khắc phục Cũng từ khóa luận nêu hạn chế mơ hình FE đƣa hƣớng nghiên cứu để cải thiện hiệu mơ hình Từ khóa: lợi nhuận ngân hàng, hiệu ngân hàng, FEM i LỜI CAM ĐOAN Khóa luận cơng trình nghiên cứu riêng tác giả, kết nghiên cứu trung thực, khơng có nội dung đƣợc cơng bố trƣớc nội dung ngƣời khác thực ngoại trừ trích dẫn đƣợc dẫn nguồn đầy đủ khóa luận Ngƣời cam đoan Lê Đức Duy ii LỜI CẢM ƠN Lời cảm ơn tác giả xin gửi đến Cô Võ Thị Ngọc Hà - Giảng viên khoa Tài Ngân hàng - Trƣờng Đại học Ngân hàng Thành phố Hồ Chí Minh Cơ ngƣời trực tiếp hƣớng dẫn khóa luận tơi suốt thời gian tơi làm khóa luận Đồng thời xin gửi lời cảm ơn đến anh, chị đáp viên yêu quý, dành thời gian quý báu tơi lời chia sẻ, nhận định thực tế vô quý giá Tôi xin gửi lời cảm ơn chân thành đến gia đình, bạn bè, động viên, giúp đỡ trình làm khóa luận Tơi xin chân thành cảm ơn! Ký tên Lê Đức Duy iii MỤC LỤC DANH MỤC TỪ VIẾT TẮT vii DANH MỤC BẢNG BIỂU VÀ BIỂU ĐỒ viii CHƢƠNG 1: GIỚI THIỆU .1 1.1 Tính cấp thiết lý chọn đề tài .1 1.2 Mục tiêu nghiên cứu 1.3 Câu hỏi nghiên cứu 1.4 Đối tƣợng phạm vi nghiên cứu 1.5 Phƣơng pháp nội dung nghiên cứu 1.6 Đóng góp đề tài 1.7 Bố cục nghiên cứu CHƢƠNG 2: CƠ SỞ LÝ THUYẾT VÀ TỔNG QUAN NGHIÊN CỨU .5 2.1 Cơ sở lý thuyết 2.1.1 Ngân hàng thƣơng mại 2.1.2 Quan niệm hiệu hoạt động ngân hàng .5 2.1.3 Các mơ hình đánh giá hiệu hoạt động ngân hàng thƣơng mại 2.1.4 2.2 Lý thuyết nhân tố ảnh hƣởng đến lợi nhuận ngân hàng Tổng quan nghiên cứu trƣớc .16 2.2.1 Các nghiên cứu nƣớc .16 2.2.2 Các nghiên cứu nƣớc .18 CHƢƠNG 3: PHƢƠNG PHÁP NGHIÊN CỨU 21 3.1 Phƣơng pháp thu thập liệu 21 3.2 Mơ hình nghiên cứu 21 3.2.1 Mơ hình nghiên cứu 21 iv 3.2.2 Mô tả biến thang đo 22 3.2.3 Giả thuyết nghiên cứu .23 3.3 Phƣơng pháp xử lý số liệu 37 3.4 Lý thuyết mơ hình 37 3.4.1 Pool Original Least Square (POLS) 37 3.4.2 Random Effect Model (REM) 39 3.4.3 Fixed Effect Model (FEM) 42 3.4.4 Ma trận trọng số 43 3.4.5 Ma trận hiệp phƣơng sai 44 3.4.6 Kiểm định phù hợp mơ hình 45 CHƢƠNG 4: KẾT QUẢ NGHIÊN CỨU VÀ THẢO LUẬN 48 4.1 Thống kê biến kết kiểm định 48 4.1.1 Thống kê biến 48 4.1.2 Kiểm định mối quan hệ tƣơng quan biến 50 4.1.3 Kiểm định đa cộng tuyến 50 4.2 Kết ƣớc lƣợng kiểm định mơ hình hồi quy ROE 51 4.2.1 Kết ƣớc lƣợng .51 4.2.2 Lựa chọn mơ hình 52 4.2.3 Kiểm định vi phạm giả định mơ hình 54 4.3 Kết ƣớc lƣợng kiểm định mơ hình hồi quy ROA 56 4.3.2 Lựa chọn mơ hình 58 4.3.3 Kiểm định khuyết tật mơ hình khắc phục 58 4.4 Thảo luận kết mơ hình .60 5.1 Kết luận 63 5.2 Hạn chế khóa luận 63 v 5.3 Hƣớng nghiên cứu 65 5.3.1 Cải thiện ƣớc lƣợng phƣơng sai 65 5.3.2 Mơ hình RE between-within 66 5.4 Đề xuất nhằm nâng cao hiệu hoạt động ngân hàng thƣơng mại Việt Nam .67 5.4.1 Đối với nhà quản lý cấp cao ngân hàng 67 5.4.2 Đối với nhà quản lý, nhà làm sách 67 TÀI LIỆU THAM KHẢO x PHỤ LỤC xiii SUMMARY xxviii vi vii DANH MỤC TỪ VIẾT TẮT Từ viết tắt BCTC FEM REM REBW H VIF NHNN NHTM NHTMCP Nguyên nghĩa tiếng anh Nguyên nghĩa tiếng việt Báo cáo tài Fixed Effect Model Mơ hình tác động cố định Random Effect Model Mơ hình tác động ngẫu nhiên Random Effect Between Hiệu ứng ngẫu nhiên Within trung bình bên Hypothesis Giả thuyết Variance inflation factor Hệ số lạm phát phƣơng sai NHNN Ngân hàng thƣơng mại Ngân hàng thƣơng mại cổ phần vii Durbin-Watson stat 1.174017 Prob(F-statistic) 0.000000 Pooled test – ROA Redundant Fixed Effects Tests Equation: EQROA Test cross-section fixed effects Effects Test Statistic Cross-section F Cross-section Chi-square d.f Prob 6.772174 138.397705 (23,279) 23 0.0000 0.0000 Cross-section fixed effects test equation: Dependent Variable: ROA Method: Panel Least Squares Date: 01/15/21 Time: 18:55 Sample: 2007 2019 Periods included: 13 Cross-sections included: 24 Total panel (balanced) observations: 312 Variable Coefficient Std Error t-Statistic Prob C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS 2.089203 0.016932 0.046512 -0.156444 -0.036663 -0.001796 -4.21E-05 0.042163 -9.48E-05 0.081713 0.798523 0.031703 0.007339 0.035438 0.002413 0.001418 0.000344 0.036484 0.004832 0.085031 2.616334 0.534088 6.337286 -4.414542 -15.19152 -1.266899 -0.122569 1.155673 -0.019616 0.960983 0.0093 0.5937 0.0000 0.0000 0.0000 0.2062 0.9025 0.2487 0.9844 0.3373 Root MSE Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 0.450027 0.995413 0.760539 1.305083 1.425051 1.353030 0.903679 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) xix 0.648741 0.638273 0.457417 63.18749 -193.5929 61.97372 0.000000 Hausman test – ROA Correlated Random Effects - Hausman Test Equation: EQROA Test cross-section random effects Test Summary Cross-section random Chi-Sq Statistic Chi-Sq d.f Prob 0.000000 1.0000 * Cross-section test variance is invalid Hausman statistic set to zero Cross-section random effects test comparisons: Variable SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS Fixed 0.186385 0.048495 -0.207091 -0.040203 -0.003253 0.000092 0.037615 0.009050 0.106526 Random Var(Diff.) Prob 0.093053 0.046890 -0.197333 -0.039068 -0.002853 -0.000049 0.034378 0.003416 0.073398 0.000944 0.000003 0.000088 0.000000 0.000000 0.000000 0.000014 0.000004 0.000278 0.0024 0.3372 0.2969 0.0827 0.0227 0.0455 0.3867 0.0049 0.0468 Cross-section random effects test equation: Dependent Variable: ROA Method: Panel Least Squares Date: 01/15/21 Time: 18:56 Sample: 2007 2019 Periods included: 13 Cross-sections included: 24 Total panel (balanced) observations: 312 Variable Coefficient Std Error t-Statistic Prob C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS -0.737897 0.186385 0.048495 -0.207091 -0.040203 -0.003253 9.25E-05 0.037615 0.009050 0.106526 1.081844 0.050744 0.007145 0.036041 0.002444 0.001248 0.000313 0.031384 0.004953 0.076270 -0.682074 3.673059 6.787250 -5.746034 -16.45216 -2.605822 0.295754 1.198526 1.827133 1.396694 0.4958 0.0003 0.0000 0.0000 0.0000 0.0097 0.7676 0.2317 0.0687 0.1636 Effects Specification Cross-section fixed (dummy variables) Root MSE Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 0.360509 0.995413 0.760539 1.008936 1.404831 1.167163 1.408242 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) xx 0.774585 0.748731 0.381233 40.54952 -124.3941 29.95993 0.000000 Phụ lục 3: Kiểm định phƣơng sai thay đổi mơ hình FE ROE ROA Heteroskedasticity test - ROE Panel Heteroskedasticity LR Test Equation: EQROE Specification: ROE C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS Null hypothesis: Residuals are homoskedastic Likelihood ratio Value 61.58245 df 24 Value -863.0375 -832.2463 df 279 279 Probability 0.0000 LR test summary: Restricted LogL Unrestricted LogL Restricted Test Equation: Dependent Variable: ROE Method: Panel Least Squares Date: 01/15/21 Time: 19:01 Sample: 2007 2019 Periods included: 13 Cross-sections included: 24 Total panel (balanced) observations: 312 Variable Coefficient Std Error t-Statistic Prob C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS -20.76337 2.473318 -0.190749 -1.673513 -0.341574 -0.006727 -0.000267 0.876843 0.231258 2.372650 11.54340 0.541441 0.076239 0.384559 0.026074 0.013321 0.003336 0.334874 0.052848 0.813808 -1.798722 4.568030 -2.502002 -4.351773 -13.10024 -0.504952 -0.079974 2.618424 4.375880 2.915491 0.0731 0.0000 0.0129 0.0000 0.0000 0.6140 0.9363 0.0093 0.0000 0.0038 Effects Specification Cross-section fixed (dummy variables) Root MSE Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 3.846668 10.67667 7.575400 5.743830 6.139725 5.902057 1.174017 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) xxi 0.741326 0.711657 4.067803 4616.620 -863.0375 24.98682 0.000000 Heteroskedasticity test - ROA Panel Heteroskedasticity LR Test Equation: EQROA Specification: ROA C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS Null hypothesis: Residuals are homoskedastic Likelihood ratio Value 114.8134 df 24 Value -124.3941 -66.98737 df 279 279 Probability 0.0000 LR test summary: Restricted LogL Unrestricted LogL Restricted Test Equation: Dependent Variable: ROA Method: Panel Least Squares Date: 01/15/21 Time: 19:03 Sample: 2007 2019 Periods included: 13 Cross-sections included: 24 Total panel (balanced) observations: 312 Variable Coefficient Std Error t-Statistic Prob C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS -0.737897 0.186385 0.048495 -0.207091 -0.040203 -0.003253 9.25E-05 0.037615 0.009050 0.106526 1.081844 0.050744 0.007145 0.036041 0.002444 0.001248 0.000313 0.031384 0.004953 0.076270 -0.682074 3.673059 6.787250 -5.746034 -16.45216 -2.605822 0.295754 1.198526 1.827133 1.396694 0.4958 0.0003 0.0000 0.0000 0.0000 0.0097 0.7676 0.2317 0.0687 0.1636 Effects Specification Cross-section fixed (dummy variables) Root MSE Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 0.360509 0.995413 0.760539 1.008936 1.404831 1.167163 1.408242 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) xxii 0.774585 0.748731 0.381233 40.54952 -124.3941 29.95993 0.000000 Phụ lục 4: Kết hồi quy khắc phục mơ hình FE với ROE ROA ROE: Dependent Variable: ROE Method: Panel EGLS (Cross-section weights) Date: 01/15/21 Time: 19:05 Sample: 2007 2019 Periods included: 13 Cross-sections included: 24 Total panel (balanced) observations: 312 Iterate weights to convergence White diagonal standard errors & covariance (d.f corrected) Convergence not achieved after weight iterations Variable Coefficient Std Error t-Statistic Prob C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS -17.79874 2.379677 -0.202592 -2.249397 -0.320786 -0.006160 0.000267 0.656503 0.212583 1.803032 11.00159 0.529368 0.061523 0.386519 0.023755 0.016707 0.002120 0.292772 0.041579 0.775944 -1.617834 4.495317 -3.292953 -5.819629 -13.50415 -0.368718 0.126105 2.242369 5.112781 2.323664 0.1068 0.0000 0.0011 0.0000 0.0000 0.7126 0.8997 0.0257 0.0000 0.0209 Effects Specification Cross-section fixed (dummy variables) Weighted Statistics Root MSE Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 3.801687 11.51595 8.011688 5.546451 5.942345 5.704678 1.228417 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) 0.768050 0.741446 4.020236 4509.280 -832.2463 28.87018 0.000000 Unweighted Statistics R-squared Sum squared resid 0.736270 4706.850 Mean dependent var Durbin-Watson stat xxiii 10.67667 1.109083 ROA: Dependent Variable: ROA Method: Panel EGLS (Cross-section weights) Date: 01/15/21 Time: 19:06 Sample: 2007 2019 Periods included: 13 Cross-sections included: 24 Total panel (balanced) observations: 312 Iterate weights to convergence White diagonal standard errors & covariance (d.f corrected) Convergence not achieved after weight iterations Variable Coefficient Std Error t-Statistic Prob C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS -0.227684 0.163315 0.041106 -0.242590 -0.037751 -0.003209 0.000176 0.023793 0.007374 0.064270 0.927170 0.045367 0.006850 0.028032 0.001971 0.000965 0.000146 0.019573 0.004041 0.052684 -0.245569 3.599877 6.000933 -8.653973 -19.15005 -3.323472 1.199521 1.215642 1.824891 1.219908 0.8062 0.0004 0.0000 0.0000 0.0000 0.0010 0.2313 0.2251 0.0691 0.2235 Effects Specification Cross-section fixed (dummy variables) Weighted Statistics Root MSE Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 0.354504 1.196201 0.908947 0.640945 1.036839 0.799172 1.259391 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) 0.837823 0.819222 0.374883 39.20998 -66.98737 45.04183 0.000000 Unweighted Statistics R-squared Sum squared resid 0.769563 41.45297 Mean dependent var Durbin-Watson stat xxiv 0.995413 1.342472 Phụ lục 5: Kiểm định phƣơng sai thay đổi mơ hình khắc phục ROE ROA ROE: Panel Heteroskedasticity LR Test Equation: EQROE Specification: ROE C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS Null hypothesis: Residuals are homoskedastic Likelihood ratio Value 61.58245 df 24 Value -863.0375 -832.2463 df 279 279 Probability 0.0000 LR test summary: Restricted LogL Unrestricted LogL Restricted Test Equation: Dependent Variable: ROE Method: Panel Least Squares Date: 01/15/21 Time: 19:10 Sample: 2007 2019 Periods included: 13 Cross-sections included: 24 Total panel (balanced) observations: 312 White diagonal standard errors & covariance (d.f corrected) Variable Coefficient Std Error t-Statistic Prob C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS -20.76337 2.473318 -0.190749 -1.673513 -0.341574 -0.006727 -0.000267 0.876843 0.231258 2.372650 12.19040 0.584935 0.068244 0.432946 0.027186 0.019773 0.002416 0.347374 0.051145 0.853005 -1.703256 4.228363 -2.795110 -3.865407 -12.56449 -0.340185 -0.110400 2.524206 4.521578 2.781520 0.0896 0.0000 0.0055 0.0001 0.0000 0.7340 0.9122 0.0122 0.0000 0.0058 Effects Specification Cross-section fixed (dummy variables) Root MSE Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 3.846668 10.67667 7.575400 5.743830 6.139725 5.902057 1.174017 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) xxv 0.741326 0.711657 4.067803 4616.620 -863.0375 24.98682 0.000000 ROA: Panel Heteroskedasticity LR Test Equation: EQROA Specification: ROA C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS Null hypothesis: Residuals are homoskedastic Likelihood ratio Value 114.8134 df 24 Value -124.3941 -66.98737 df 279 279 Probability 0.0000 LR test summary: Restricted LogL Unrestricted LogL Restricted Test Equation: Dependent Variable: ROA Method: Panel Least Squares Date: 01/15/21 Time: 19:12 Sample: 2007 2019 Periods included: 13 Cross-sections included: 24 Total panel (balanced) observations: 312 White diagonal standard errors & covariance (d.f corrected) Variable Coefficient Std Error t-Statistic Prob C SIZE EQ DPRR CIR HHIDR TTTD GDP CPI CRISIS -0.737897 0.186385 0.048495 -0.207091 -0.040203 -0.003253 9.25E-05 0.037615 0.009050 0.106526 1.347546 0.065991 0.009398 0.046295 0.002703 0.001099 0.000187 0.029798 0.005254 0.101387 -0.547586 2.824382 5.160446 -4.473340 -14.87571 -2.961424 0.495717 1.262328 1.722460 1.050685 0.5844 0.0051 0.0000 0.0000 0.0000 0.0033 0.6205 0.2079 0.0861 0.2943 Effects Specification Cross-section fixed (dummy variables) Root MSE Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 0.360509 0.995413 0.760539 1.008936 1.404831 1.167163 1.408242 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) xxvi 0.774585 0.748731 0.381233 40.54952 -124.3941 29.95993 0.000000 Phụ lục 6: Các ngân hàng phạm vi nghiên cứu luận văn Tên ngân hàng Ngân hàng TMCP Á Châu Ngân hàng TMCP An Bình Ngân hàng Nơng nghiệp Phát triển Nông thôn Việt Nam Ngân hàng TMCP Đầu tƣ Phát triển Việt Nam Ngân hàng TMCP Công Thƣơng Việt Nam Ngân hàng TMCP Xuất nhập Việt Nam Ngân hàng TMCP Bản Việt (đã đổi tên từ Ngân hàng TMCP Gia Định) Ngân hàng TMCP Phát triển TP.HCM Ngân hàng TMCP Kiên Long Ngân hàng TMCP Quân Đội Ngân hàng TMCP Hàng hải Việt Nam Ngân hàng TMCP Nam Á Ngân hàng TMCP Quốc Dân Ngân hàng TMCP Phƣơng Đơng Ngân hàng TMCP Sài Gịn Cơng Thƣơng Ngân hàng TMCP Sài Gịn - Hà Nội Ngân hàng TMCP Sài Gòn Ngân hàng TMCP Sài Gòn Thƣơng Tín Ngân hàng TMCP Kỹ thƣơng Việt Nam Ngân hàng TMCP Việt Á Ngân hàng TMCP Ngoại thƣơng Việt Nam Ngân hàng TMCP Quốc tế Việt Nam Ngân hàng TMCP Việt Nam Thịnh Vƣợng Ngân hàng TMCP Xăng Dầu Petrolimex Tên viết tắt ACB ABB Agribank BID CTG EIB BVB HDB KLB MBB MSB NAB NVB OCB SGB SHB SCB STB TCB VietABank VCB VIB VPB PGB xxvii SUMMARY Theories and previous studies One of the important components of the financial system and the economy is commercial banks In recent years, commercial banks have made a great contribution to the financial development of the country and the world economy The financial performance of the banks affects the capital allocation, business expansion, the economic growth of industries and the development of the economy Therefore, the profitability of banks not only affects commercial banks, but it also affects macroeconomics Since then, profits become an important part of the operations of banks and affect many areas The goal of the study is to find a suitable model to consider factors affecting the banking efficiency as well as the limitations of that model, then propose the next research directions In general, two theories commonly used to evaluate banking performance are structural efficiency theory and market power theory According to Jeon and Miller (2005): The efficient structural theory includes two hypotheses: the efficiency hypothesis X and the scale efficiency hypothesis; Market power theory consists of two hypotheses: the traditional structural efficiency hypothesis and management and the relative market power hypothesis The previous studies can be mentioned as: Ash Demirguc-Kunt and Harry Huizinga (1999) analyzed banks in 80 countries in the period 1988-1995 with a maximum total of 7900 commercial banks Although the author uses the POLS model, the big observations have filled most of the problems of the model; Kosmidou et al (2005) who studied 32 banks in the UK between 1995 and 2002 using the FEM model, came up with the conclusions about the equity ratio, the ratio of total operating expenses, bank size, ; Alper and Anbar (2011) studied the performance of Turkish banks in the period 2002-2010 The results of the study show that asset size and non-interest income have a positive and significant effect on a bank's profitability Research Methods xxviii The research object is the business performance of commercial banks Specifically, research on the effectiveness of profits from joint stock commercial banks listed on the Ho Chi Minh Stock Exchange (HSX), the Hanoi Stock Exchange (HNX) and the UpCom stock exchange in addition to commercial banks Others include 24 joint stock commercial banks The scope of the study is the period 2007-2019 The thesis will measure the bank's performance by ROE and ROA ratios along with other impact variables which will be outlined below along with the research model: Yit  ci    k X it ,k  uit k 1 Inside: ● Yit shows the performance of the ith bank in year t ● k is the kth separate regression coefficient of the kth impact factor with k = 1, 2, 3,…, ● X it , k is the kth variable that affects banking performance k = 1, 2, 3,…, ● At k = 1,2,…, 8, X it ,k are variable asset size; equity ratio; variables of credit risk provisions; operating cost ratio; income diversification; credit growth; GDP growth; inflation index at bank i and year t ● At k = 9, X it ,9 is the dummy variable for the economic crisis in 2009-2010 ● Uit is the error of the model Before the regression study, the thesis will perform basic tests on correlation test and multicollinearity to consider whether independent variables are suitable for regression analysis or not The results obtained are very satisfactory for the implementation of the regression model (results in the appendix) The thesis will study three models including POLS, FEM, REM for both dependent variables, ROE and ROA After the regression results are available, the appropriate model will be selected and the final tests are performed to check the common errors encountered by the regression model and correct (if any) xxix The results The results of the regression implementation for dependent variables ROE are: POLS: Yit  3.9  1.11 Sizeit 0.23 Eqit 1.12 DPRRit  0.34 CIR it  0.006 HHIDRit (0.66) (0.00) (0.00) (0.00) (0.00) (0.72)  0.0026 TTTDit  0.84 GDPit  0.16 CPI it  CRISISt (0.50) (0.04) (3.1) (0.04) (0.004) FEM: Yit  20.76 2.47 Sizeit 0.19 Eqit 1.67 DPRRit  0.34 CIR it  0.006 HHIDRit (0.07) (0.00) (0.01) (0.00) (0.00) (0.61)  0.0003TTTDit  0.88 GDPit  0.23 CPIit  2.37 CRISISt (0.93) (0.01) (0.00) (3.2) (0.00) REM: Yit  10.17 1.91 Sizeit 0.20 Eqit  1.59 DPRRit  0.34 CIR it  0.004 HHIDRit (0.31) (0.00) (0.01) (0.00)  0.001TTTDit  0.84 GDPit  0.20 CPIit  2.14 CRISISt (0.69) (0.01) (0.00) (0.00) (0.00) (0.74) (3.3) Doing a pooled test and getting results that the impact factor does not change over time which means that FEM or REM will be more suitable than POLS (test results are in the appendix) Next, perform a Hausman test to choose between FEM and REM However, the test did not give any meaningful results However, when considering the hypothesis theory to establish the Hausman test, there are a few things to note The assumption of a Hausman test is based on two strict REM assumptions Achieving REM consistency requires adding non-correlation between the general regression coefficient and the residual making the assumption more stringent If compared with the assumption of FEM, the FEM assumption allows the correlation between the general regression coefficient and the residual, in econometric models this means that other economic factors cannot observe the correlation This is relative to the model remainder and this is a common occurrence in quantitative models because of the problem of missing variables However, with the fixed effects method, the estimated coefficient is still a consistent and unbiased estimate In addition, to achieve efficiency (more than FEM), REM is difficult to calculate the residual variance so it makes more stringent assumptions to perform the calculation, but it must be stressed that this assumption is a assumes standard support (considered very strict) and it is not tested by the Hausman test If this assumption is violated, the Hausman test will have a non-standard limit distribution, which in turn leads to erroneous results And in fact this assumption is very easy to be violated xxx when the remainder often have heterogeneous variance or autocorrelation In addition, the difference between the experimental FEM and REM is that the general regression coefficient of FE is significant while RE is not, and after checking the stationary effects of the unobserved fixed effects of each cross unit, the on the chain effect, the value of ci is really meaningful Therefore, the thesis concludes that the fixed effects model (FEM) is more suitable than the random effect model (REM) in terms of general theory and research objectives Finally, check the model defects is dependent correlation and variance change of residue (results in appendix) The results show that there is no dependent correlation but still show that variance does exist The thesis uses "weight crosssection" weight matrix and calculates the White diagonal covariance matrix to overcome the limitations of the model (regression results in the appendix) But the results still show that the variance exists Perform the same as for ROE, regression results show that FEM is the most suitable for the dependent variable ROA (other regression results in appendix) FEM: Yit  0.74 0.19 Sizeit 0.05 Eqit  0.21 DPRRit  0.04 CIR it  0.003 HHIDRit (0.50) (0.00) (0.00) (0.00) (0.00)  0.0001TTTDit  0.04 GDPit  0.0091CPIit  0.11CRISISt (0.77) (0.23) (0.07) (0.16) (0.01) (3.4) The results of the model defect test received results similar to the FE regression for ROE and the corrected regression results still exist variable variance although no longer dependent correlation (regression results in appendix) Conclusion, limitations and next research directions The experimental results show that out of independent variables, selected thesis shows the impact on the bank's operational efficiency through ROE and ROA indicators Variables having a positive impact are banking size, GDP growth, inflation, and the period of crisis 2009-2010 The variables that have the opposite effect are the equity to total capital ratio, risk reserve ratio, and total operating expenses The FE model was then tested to check and showed no problem of cocorrelation of variance but the appearance of variance was changed So the thesis has implemented corrective measures to solve the problem of variance change, but the model still exists for variable variance xxxi The next paragraph of the thesis presents the limitations and next research directions to improve the efficiency of the model The FE model has five main limitations First, the regression coefficients not fully describe the explanatory variable information Second, estimating residual variance is not "correct" Third, R-squared reduced reliability Fourth, the fixed effects model (FE) is not as effective as the random effects model (RE) Fifth, the general regression coefficient (unobserved fixed effects vary by cross unit) are often misinterpreted The thesis proposes two further research directions to improve the model (especially the effectiveness of the model) Thesis suggests methods to estimate grouped variance (Cluster-Robust Covariances) Observations are then grouped into different groups where the residues are correlated with other observations in the same group and not with observations in different groups Specifically, the thesis suggests two firm group variance estimates Firstly, White cross-section, this method assumes that the remainders are correlated between the cross units (correlated simultaneously) with the groups divided by time T thereby creating variance of variation Secodly, White period, this method assumes that the remainder of the same cross unit at different times are correlated and also exist autocorrelated by grouping the N units together In many comparisons between the fixed-effects model and the random-effects model, the FE model is frequently chosen for its bias-limiting properties with much less stringent assumptions than RE However, the FE model also brings limitations as mentioned above So the thesis proposes a new model that allows the bias not only to be limited but also to be modeled clearly The model uses regression within cross units such as the method of FE (within regression) and between regression (between regression) The model called RE between-within (REBW) is regressed by the direction of the GLS classes, and the model has a simple form:   Yit  ci  1 ( X it  X i )   X i  uit (4.1) xxxii xxxiii ...NGÂN HÀNG NHÀ NƢỚC VIỆT NAM BỘ GIÁO DỤC VÀ ĐÀO TẠO TRƢỜNG ĐẠI HỌC NGÂN HÀNG TP HỒ CHÍ MINH - ĐỀ CƢƠNG KHÓA LUẬN CÁC NHÂN TỐ ẢNH HƢỞNG ĐẾN HIỆU QUẢ HOẠT ĐỘNG CỦA NGÂN HÀNG... từ nhà nƣớc nhà nƣớc góp phần quan trọng phục hồi hiệu hoạt động ngành ngân hàng Do khóa luận kết luận quy mơ ngân hàng tác động chiều đến hiệu hoạt động ngân hàng 24 Đồ thị 3.1: So sánh hiệu. .. lý thuyết nhân tố tác động đến hiệu hoạt động ngân hàng đƣợc trình bày chƣơng hai, khóa luận đƣa giả định ảnh hƣởng 10 biến nhân tố đến hoạt động kinh doanh ngân hàng ● H1: Quy mô ngân hàng (Size)