Luận án Nghiên cứu chất lượng thông tin lợi nhuận kế toán công bổ của các ngân hàng thương mại Việt Nam

growth 0.091762 0.038304 2.396 0.0179 * ## ovh -0.832518 2.388362 -0.349 0.7279 ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 0.1 ' '1 ## ## Residual standard error: 0.1197 on 146 degrees of freedom ## (27 observations deleted due to missingness) ## Multiple R-squared: 0.5633, Adjusted R-squared: 0.5453 ## F-statistic: 31.38 on 6 and 146 DF, p-value: < 2.2e-16 vif(cpm) ## cfo ebt deposit loan growth ovh ## 1.453234 1.123167 1.543974 1.602391 1.744049 1.601306 bgtest(cpm) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: cpm ## LM test = 10.342, df = 1, p-value = 0.0013 dwtest(cpm,alternative = "two.sided") ## Durbin-Watson test ## data: cpm ## DW = 2.2995, p-value = 0.09187 ## alternative hypothesis: true autocorrelation is not 0 bptest(cpm) ## studentized Breusch-Pagan test ## data: cpm ## BP = 8.2971, df = 6, p-value = 0.2171 bptest(cpm,varformula=~fitted.values(cpm),studentize=TRUE,data=CashpredicP) ## studentized Breusch-Pagan test ## data: cpm ## BP = 1.6986, df = 1, p-value = 0.1925 resettest(c pm,type="fitted") ## RESET test ## data: cpm ## RESET = 0.76074, dfl = 2, df2 = 144, p-value = 0.4692 RELAIMPO - CASH PREDICTION MODEL cpm.rela<-calc.relimp(cpm) print(cpm.rela) ## Response variable: acfol ## Total response variance: 0.0315031 ## Analysis based on 153 observations ## 6 Regressors: ## Img ## cfo 0.480689565 ## ebt 0.005118003 ## deposit 0.002949513 ## loan 0.021700723 ## growth 0.034642170 ## ovh 0.018174575 ## Average coefficients for different model sizes: ## ## IX 2Xs 3Xs 4Xs ## cfo -1.2575751 -1.291658178 -1.32575745 -1.35677999 -1 ## ebt 0.9668024 1.359068845 1.66051427 1.80758364 1 ## deposit 0.1306306 0.006620732 -0.04989937 -0.05728476 -0 ## loan 0.2994243 0.231076295 0.19704576 0.17659859 0 ## growth -0.1352461 -0.079651727 -0.03099040 0.01282859 0 ## ovh 8.2057841 5.346253008 3.07948431 1.31415034 0 ## 6Xs ## cfo -1.407879705 ## ebt 1 793768888 ## deposit 0.007009316 ## loan 0.130359867 ## growth 0.091762433 ## ovh -0.832517524 ## cfo ebt deposit loan growth ovh ## Proportion of variance explained by model: 56.33% ## Metrics are not normalized (rela=FALSE). ## Relative importance metrics: 5Xs .384155020 .835295596 .034489318 .156564916 .053435464 006239135 plot(cpm.rela) Relative importances for acfol Method LMG o £ s n — ro o s - UỈ 1 J c o - GO _ <D o _ ° o - I I I I I o’- cfo ebt depo loan grow ovh R2 = 5633%, metrics are not normalized. KẾT QUẢ HÔI QUY CHO GIAI ĐOẠN 1: TỪ2008 - 2011 cpml <-lm(acfol~cfo+ebt+deposỉt+loan+growth+ovh,data=CashpredỉcPl) summary(cpml) ## Call: ## lm(formula = acfol ~ cfo + ebt + deposit + loan + growth + ovh, ## data = CashpredỉcPl) ## ## Residuals: ## Min IQ Median 3Q Max ## -0.24988 -0.09022 -0.00797 0.07727 0.51197 ## ## Coefficients: ## Estimate std. Error t value Pr(>|t1) ## (Intercept) -0.05852 0.09709 -0.603 0.548 ## cfo -1.48863 0.19891 -7.484 7.73e-ll *** ## ebt 0.73610 2.19890 0.335 0 739 ## deposit 0.20632 0 17857 1.155 0.251 ## loan 0.08464 0.17658 0.479 0.633 ## growth 0.06892 0.05235 1.316 0.192 ## ovh -3.67793 4.76292 -0.772 0.442 ## — ## Signif. codes: 0 '***' 0.001 '** ' 0.01 0.05 '.' 0.1 ## ## Residual standard error: 0.1442 on 81 degrees of freedom ## Multiple R-squared: 0.4958, Adjusted R-squared: 0.4585 ## F-statistic: 13.28 on 6 and 81 DF, p-value: 2.048e-10 vif(cpml) ## cfo ebt deposit loan growth ovh ## 1.553949 1.069025 1.707157 2.541650 1.517749 1.865223 bgtest(cpml) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: cpml ## LM test = 2.9235, df = 1, p-value = 0.0873 dwtest(cpml,alternative = "two.sided") ## Durbin-Watson test ## data: cpml ## DW = 2.2281, p-value = 0.3642 ## alternative hypothesis: true autocorrelation is not 0 bptest(cpml) ## studentized Breusch-Pagan test ## data: cpml ## BP = 1.6465, df = 6, p-value = 0.9492bptest(cpml,varformula = ~ fitted.values(cpml),studentize = TRUE,data=Cashp redicPl) ## studentized Breusch-Pagan test ## data: cpml ## BP = 0.14478, df = 1, p-value = 0.7036 resettest(cpml,type="fitted") ## RESET test ## data: cpml ## RESET = 0.61835, dfl = 2, df2 = 79, p-value = 0.5414 RELAIMPO - CASH PREDICTION MODEL 2008-2011 cpml.rela<-calc.relimp(cpml) print(cpml.rela) ## Response variable: acfol ## Total response variance: 0.03839251 ## Analysis based on 88 observations ## ## 6 Regressors: ## cfo ebt deposit loan growth ovh ## Proportion of variance explained by model: 49.58% ## Metrics are not normalized (rela=FALSE). ## ## Relative importance metrics: ## ## Img ## cfo 0.393568212 ## ebt 0.001189728 ## deposit 0.012129597 ## loan 0.049795203 ## growth 0.024246406 ## ovh 0.014914049 ## ## Average coefficients i for different model sizes: ## ## IX 2Xs 3Xs 4Xs 5Xs ## cfo -1.3817829 - 1.39107896 -1.41058563 -1.43520680 -1.46191397 ## ebt 1.2748769 0.74084781 0.73618137 0.82054009 0 87098856 ## deposit 0 2933768 0.14288665 0.09062836 0.09705649 0.13745253 ## loan 0.4988327 0.41472281 0.34022726 0.26731644 0.18655234 ## growth -0.1313544 - 0.07153283 -0.02697240 0.00815250 0.03858959 ## ovh 9.9335461 4.56998126 0.64605058 -2.06110774 -3.58890100 ## 6Xs ## cfo -1.48862574 ## ebt 0 73609788 ## deposit 0.20631682 ## loan 0.08463802 ## growth 0.06891587 ## ovh -3.67792987 plot(cpml.rela) Relative importances for acfo! Method LMG R2 = 49.58%, metrics are not normalized. KÉT QUẢ HÒI QUY CHO GIAIĐOẠN2: TỪ2012 ĐÉN 2015 cpm2 <-lm(acfol~cfo+ebt+deposit+loan+growth+ovh,data=CashpredicP2) summary(cpm2) ## Call: ## lm(formula = acfol ~ cfo + ebt + deposit + loan + growth + ovh, ## data = CashpredicP2) ## ## ## Residuals: 3Q Max Min IQ Median ## -0.169623 -0 .047860 0.000522 0. 042350 0.191839 ## ## Coefficients I ## Estimate std. Error t value Pr(>|t|) ## (Intercept) -0.04158 0.06394 -0.650 0.518089 ## cfo -1.42435 0.10690 - 13.324 < 2e-16 *** ## ebt 1.08512 1.73528 0.625 0.534208 ## deposit -0 15787 0.10406 -1.517 0.134680 ## loan 0.10943 0.08699 1.258 0.213437 ## growth 0.19902 0.05664 3.514 0.000864 *** ## ovh 3.78459 1.96840 1.923 0.059437 . ## — ## Signif. codes: 0 '***' 0.001 '** ' 0.01 0.05 '.' 0.1 ## ## Residual standard error: 0.07128 on 58 degrees of freedom ## (23 observations deleted due to missingness) ## Multiple R-squared: 0.7936, Adjusted R-squared: 0.7722 ## F-statistic: 37.16 on 6 and 58 DF, p-value: < 2.2e-16 vif(cpm2) ## cfo ebt deposit loan growth ovh ## 1.448983 1.340653 1.339480 1.593809 1.477848 1.335115 bgtest(cpm2) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: cpm2 ## LM test = 1.2825, df = 1, p-value = 0.2574 dwtest(cpm2,alternative = "two.sided") ## Durbin-Watson test ## data: cpm2 ## DW = 2.2469, p-value = 0.4435 ## alternative hypothesis: true autocorrelation is not 0 bptest(cpm2) ## studentized Breusch-Pagan test ## data: cpm2 ## BP = 3.7559, df = 6, p-value = 0.7097 bptest(cpm2,varformula = ~fitted.values(cpm2),studentize = TRUE,data=Cashpr edicP2) ## studentized Breusch-Pagan test ## data: cpm2 ## BP = 0.013571, df = 1, p-value = 0.9073 resettest(c pm2,type="fitted") ## RESET test ## data: cpm2 ## RESET = 0.067232, dfl = 2, df2 = 56, p-value = 0.9351 RELAIMPO - CASH PREDICTION MODEL 2012 - 2015 cpm2.rela<-calc.relimp(cpm2) print(cpm2.rela) ## Response variable: acfol ## Total response variance: 0.022307 ## Analysis based on 65 observations ## ## 6 Regressors: ## cfo ebt deposit loan growth ovh ## Proportion of variance explained by model: 79.36% ## Metrics are not normalized (rela=FALSE). ## ## Relative importance metrics: ## Img ## cfo 0.671574688 ## ebt 0.011012268 ## deposit 0.016819366 ## loan 0.004902236 ## growth 0.056183996 ## ovh 0.033070366 ## ## Average coefficients for different model sizes: ## ## IX 2Xs 3Xs 4Xs 5Xs ## cfo -1.26847252 -1.29734387 -1.328094517 -1.35925120 -1.39109007 ## ebt 3.53621119 3.28863901 2.970366345 2.49647051 1.87486416 ## deposit -0.21781722 -0.22512323 -0.209103119 -0.18866459 -0.16988444 ## loan -0.02411531 -0.01817961 0.006932275 0.03971601 0.07411951 ## growth -0 22828367 -0 13927796 -0.049248964 0.03707929 0.11944533 ## ovh 7.20204116 6.25656675 5.239759488 4.40416695 3.88864576 ## 6Xs ## cfo -1.4243513 ## ebt 1.0851217 ## deposit -0.1578661 ## loan 0.1094263 ## growth 0.1990167 ## ovh 3.7845900 plot(cpm2.rela) Relative importances for acfoi Method LMG <D o c ra ra > <D tn C O <n q> L O ...p O- cfo ebt depo loan grow ovh 2 i - i 1- I R = 79.36%, metrics are not normalized. PHẦN 3: KẾT QUẢ HỒI QUY DỮ LIỆU BẢNG - PANEL DATA library(plm) ## Loading required package: Formula pcash<-pirn.data(CashpredicP,indexes=c("bank","year")) MÔ HÌNH TÁC ĐỘNG Gộp (POOLING MODEL) cmpool<-plm(acfol~cfo+ebt+loan+deposỉt+growth+ovh,data=pcash,model="pooling ") z \ summary(cmpool) ## Oneway (individual) effect Pooling Model ## Call: ## plm(formula = acfol ~ cfo + ebt + loan + deposit + growth + ovh, ## data = pcash, model = "pooling") ## ## Unbalanced Panel: n=22, T=6-7, N=153 ## Residuals : ## Min. 1st Qu. Median 3rd Qu. Max. ## -0.2730 -0. 0657 -0.0165 0.0528 0 .5240 ## ## Coefficients : ## Estimate ! std. Error t-value Pr(>|t|) ## (Intercept) -0.0555261 0.0615477 -0.9022 0.36846 ## cfo -1.4078797 0.1144966 -12.2963 < 2e-16 *** ## ebt 1 7937689 1.3811463 1 2988 0.19608 ## loan 0.1303599 0.0910340 1.4320 0.15428 ## deposit 0.0070093 0.1006590 0.0696 0.94458 ## growth 0.0917624 0.0383042 2.3956 0.01786 * ## ovh -0.8325175 2.3883623 -0.3486 0.72791 ## — ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ## ## Total Sum of Squares: 4.7885 ## Residual Sum of Squares : 2.0912 ## R-Squared: 0.56327 ## Adj. R-Squared: 0.5375 ## F-statistic: 31.3844 on 6 and 146 DF, p-value: < 2.22e-16 DIAGNOSTIC TEST FOR POOLING MODEL pdwtest(cmpool) ## Durbin-Watson test for serial correlation in panel models ## data: acfol ~ cfo + ebt + loan + deposit + growth + ovh ## DW = 2.2995, p-value = 0.9541 ## alternative hypothesis: serial correlation in idiosyncratic errors pbgtest(cmpool) ## Breusch-Godfrey/Wooldridge test for serial correlation in panel models ## data: acfol ~ cfo + ebt + loan + deposit + growth + ovh ## chisq = 14.612, df = 6, p-value = 0.0235 ## alternative hypothesis: serial correlation in idiosyncratic errors bptest(cmpool) ## studentized Breusch-Pagan test ## data: cmpool ## BP = 8.2971, df = 6, p-value = 0.2171 MÔ HÌNH TÁC ĐỘNG CÓ ĐỊNH- FIXED EFFECT MODEL (FEM) cmfe<-plm(acfol~cfo+ebt+loan+deposit+growth+ovh,model="within",data=pcash) summary(cmfe) ## Oneway (individual) effect Within Model ## Call: ## plm(formula = acfol ~ cfo + ebt + loan + deposit + growth + ovh, ## data = pcash, model = "within") ## ## Unbalanced Panel: n=22, T=6-7, N=153 ## ## Residuals : ## Min. 1st Qu. Median 3rd Qu. Max. ## -0.2830 -0.0529 -0.0014 0.0549 0.4180 ## ## Coefficients : ## Estimate std. Error t-value Pr(>|t|) ## cfo -1.377109 0.115318 -11.9418 < 2.2e-16 *** ## ebt 3.018402 1.730821 1.7439 0.08363 • ## loan 0.625820 0.146440 4.2736 3.779e-05 *** ## deposit -0.188805 0.133845 -1.4106 0.16084 ## growth 0.100388 0.042214 2 3781 0.01892 * ## ovh 3.607522 3 327037 1.0843 0 28032 ## ... ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 0.1 ' '1 ## ## Total Sum of Squares: 4.6996 ## Residual Sum of Squares: 1.6532 ## R-Squared: 0.64823 ## Adj. R-Squared: 0.5296 ## F-statistic: 38.3907 on 6 and 125 DF, p-value: < 2.22e-16 DIAGNOSTIC TEST FOR FIXED MODEL pdwtest(cmfe) ## Durbin-Watson test for serial correlation in panel models ## data: acfol ~ cfo + ebt + loan + deposit + growth + ovh ## DW = 2.3709, p-value = 0.9898 ## alternative hypothesis: serial correlation in idiosyncratic errors pbgtest(cmfe) ## Breusch-Godfrey/Wooldridge test for serial correlation in panel models ## data: acfol ~ cfo + ebt + loan + deposit + growth + ovh ## chisq = 31.82, df = 6, p-value = 1.766e-05 ## alternative hypothesis: serial correlation in idiosyncratic errors bptest(cmfe) ## studentized Breusch-Pagan test ## data: cmfe ## BP = 8.2971, df = 6, p-value = 0.2171 MÔ HÌNH TÁC ĐỘNG NGẪU NHIÊN - RANDOM EFFECT MODEL cmre<-plm(acfol~cfo+ebt+loan+deposit+growth+ovh,model="random",data=pcash) summary(cmre) ## Oneway (individual) effect Random Effect Model ## (Swamy-Arora's transformation) ## ## Call: ## plm(formula = acfol ~ cfo + ebt + loan + deposit + growth + ovh, ## data = pcash, model = "random") ## ## Unbalanced Panel: n=22, T=6-7, N=153 ## ## Effects: ## Warning in sqrt(sigma2): NaNs produced ## var std.dev share ## idiosyncratic 0.013225 0.115002 1.107 ## individual -0.001279 NA -0.107 ## theta : ## Min. 1st Qu. Median Mean 3rd Qu. Max. ## -0.7588 -0.7588 -0.7588 -0.7503 -0.7588 -0.5431 ## ## Residuals : ## Min. 1st Qu. Median Mean 3rd Qu. Max. ## -0.28900 -0.06860 -0.01430 0.00033 0.06520 0.53600 ## ## Coefficients : t-value Pr(>|t|) ## Estimate std. Error ## (Intercept) -0.026777 0.050080 -0.5347 0.593686 ## cfo -1.336053 0.117859 -11.3361 < 2.2e-16 *** ## ebt 1.450847 1.167095 1.2431 0.215813 ## loan 0.025983 0.066661 0 3898 0.697266 ## deposit 0.068793 0.082445 0.8344 0.405413 ## growth 0.099653 0 036677 2.7170 0.007384 ** ## ovh -1.551793 1.843389 -0.8418 0.401268 ## — ## Signif. codes: 0 '*** ' 0.001 '** ' 0.01 '* ' 0.05 '.' 0.1 ## ## Total Sum of Squares: 4.9688 ## Residual Sum of Squares: 2.4363 ## R-Squared: 0.50968 ## Adj. R-Squared: 0.48636 ## F-statistic: 25.2937 on 6 and 146 DF, p-value: < 2.22e-16 DIAGNOSTIC TEST FOR RANDOM MODEL pdwtest(cmre) ## Durbin-Watson test for serial correlation in panel models ## data: acfol ~ cfo + ebt + loan + deposit + growth + ovh ## DW = 2.2223, p-value = 0.8739 ## alternative hypothesis: serial correlation in idiosyncratic errors pbgtest(cmre) ## Breusch-Godfrey/Wooldridge test for serial correlation in panel models ## data: acfol ~ cfo + ebt + loan + deposit + growth + ovh ## chisq = 8.2262, df = 6, p-value = 0.222 ## alternative hypothesis: serial correlation in idiosyncratic errors bptest(cmre) ## studentized Breusch-Pagan test ## data: cmre ## BP = 8.2971, df = 6, p-value = 0.2171 CÁC KIỀM ĐỊNH LựA CHỌN MÔ HÌNH TÓI ưu FIXED EFFECT MODEL vs. POOLING MODEL (p value is small => fixed effect model) pFtest(cmfe,cmpool) ## F test for individual effects ## data: acfol ~ cfo + ebt + loan + deposit + growth + ovh ## F = 1.5773, dfl = 21, df2 = 125, p-value = 0?06514 ## alternative hypothesis: significant effects FIXED EFFECT MODEL vs. RANDOM EFFECT MODEL phtest(cmfe,cmre) ## Hausman Test ## data: acfol ~ cfo + ebt + loan + deposit + growth + ovh ## chisq = 321.1, df = 6, p-value < 2.2e-16 ## alternative hypothesis: one model is inconsistent RANDOM EFFECT vs. OLS (pvalue is small => random effect model) plmtest(cmpool) ## Lagrange Multiplier Test - (Honda) ## data: acfol ~ cfo + ebt + loan + deposit + growth + ovh ## normal = -0.8403, p-value = 0.4007 ## alternative hypothesis: significant effects PHỤ LỤC 7. KẾT QUẢ HỒI QUY MÔ HÌNH 3 - ĐIỀU CHỈNH Dự PHÒNG RỦI RO TÍN DỤNG PHẦN 1. TÍNH TOÁN CÁC BIẾN smoothing <- readXL("D:/Luan an/DRAFT/6.2016/Dulỉeu06.2016.xlsx", rownames= FALSE, header=TRUE, na="", sheet="Panel", strỉngsAsFactors=TRUE) smoothỉng$ebtllp <- with(smoothing, log(Ebtllp)) smoothỉng$llp <- wỉth(smoothing, log(Llp)) Warning in log(Llp): NaNs produc ed smoothỉng$bllp <- with(smoothing, Bllp/Bloan) smoothỉng$bllplog <- with(smoothỉng, log(Bllp)): NaNs produced smoothỉng$npl <- with(smoothing, Npl/Loan) smoothỉng$npllog <- with(smoothing, log(Npl)) smoothỉng$bloan <- with(smoothing,log(Bloan)) smoothỉng$dloan <- with(smoothing,Dloan/Asset) smoothỉng$loan <- with(smoothing, Loan/Asset) smoothỉng$loanlog <- with(smoothỉng,log(Loan)) smoothỉng$equỉty <- with(smoothing, Equỉty/Asset) smooth=2012) PHẦN 2: KẾT QUẢ HỒI QUY OLS GIAI ĐOẠN 2008 - 2015 sm <-lm(llp~bllp+ebtllp+equỉty+npllog+gdp,data=smooth) summary(sm) ## Call: ## lm(formula = lip ~ blip + ebtllp + equity + npllog + gdp, data = smooth) ## Residuals: ## Min IQ Median 3Q Max ## -2.02762 -0.30951 0.07956 0.31948 1.28924 ## Coefficients: ## Estimate std. Error t value Pr(>|t1) ## (Intercept) -3.60533 0.78925 -4.568 9.67e-06 *** ## blip 30.64969 6.66903 4.596 8.59e-06 *** ## ebtllp 0.55780 0.05069 11.004 < 2e-16 *** ## equity -0.76665 0.91132 -0.841 0.401 ## npllog 0.61796 0.05731 10 782 < 2e-16 *** ## gdp -2.66559 7.11883 -0.374 0.709 ## — ## Signif. 1 codes: 0 '***' 0.001 **' 0.01 0.05 ' . ' 0.1 ## Residual standard error: 0.525 ( an 163 degrees of freedom ## (11 observations deleted due to missingness) ## Multiple R-squared: 0.9105, Adjusted R-squared: 0.9078 ## F-statistic: 331.7 on 5 and 163 DF, p-value: < 2.2e-16 DIAGNOSTIC TEST vif(sm) ## blip ebtllp equity npllog gdp ## 1.284242 3.193828 1.715020 3.361711 1.015700 bgtest(sm) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: sm ## LM test = 0.12099, df = 1, p-value = 0.728 dwtest(sm) ## Durbin-Watson test ## data: sm ## DW = 2.0389, p-value = 0.5364 ## alternative hypothesis: true autocorrelation is greater than 0 bptest(sm) ## studentized Breusch-Pagan test ## data: sm ## BP = 2.6211, df = 5, p-value = 0.7582 bptest(sm,varformula = ~fitted.values(sm),studentize = TRUE,data=smooth) ## studentized Breusch-Pagan test ## data: sm ## BP = 2.0947, df = 1, p-value = 0.1478 resettest(sm,type="regressor") ## RESET test ## data: sm ## RESET = 0.88092, dfl = 10, df2 = 153, p-value = 0.5525 resettest(sm,type="fitted") ## RESET test ## data: sm ## RESET = 0.45078, dfl = 2, df2 = 161, p-value = 0.6379 RELAIMPO - SMOOTHING MODEL sm.rela<-calc.relimp.Im(sm) print(sm.rela) ## Response variable: lip ## Total response variance: 2.988335 ## Analysis based on 169 observations ## 5 Regressors: blip ebtllp equity npllog gdp ## Proportion of variance explained by model: 91.05% ## Metrics are not normalized (rela=FALSE). ## Relative importance metrics: # Img ## blip 0.0899983866 ## ebtllp 0.3481218662 ## equity 0.1163609371 ## npllog 0.3555267548 ## gdp 0.0005121439 ## ## Average coefficients for different model sizes: ## IX 2Xs 3Xs 4Xs 5Xs ## blip 126.932673 78.3650309 50.5215389 36.8452253 30.6496898 ## ebtllp 1.085502 0.9190972 0.7744552 0.6534125 0.5578028 ## equity -17.898679 -9.8222619 -4.7110751 -1.9116425 -0.7666545 ## npllog 1.209632 1.0440742 0 8872586 0.7437907 0.6179597 ## gdp -6 627223 -4.1790995 -3.3978813 -2 8988192 -2.6655892 plot(sm.rela) Relative importances for lip Method LMG 2 ___ . - a a- I R = 91.05%, metrics are not normalized. GIAI ĐOẠN TỪ2008 ĐẾN 2011 sml <-lm(llp~bllp+ebtllp+equity+npllog+gdp,data=smoothpl) summary(sml) ## Call: ## lm(formula = lip ~ blip + ebtllp + equity + npllog+gdp, data = smoothpl) ## Residuals: ## Min IQ Median 3Q Max ## -1.92952 -0.27423 0.07398 0.33006 1.19062 ## Coefficients: ## Estimate std. Error t value Pr(>|t1) ## (Intercept) -3.10838 1.07745 -2.885 0.005 ** ## blip 12.88129 14.68981 0.877 0.383 ## ebtllp 0.57910 0.08317 6.963 7.58e-10 *** ## equity -0.80539 1.12500 -0.716 0.476 ## npllog 0.58912 0.09600 6.136 2.85e-08 *** ## gdp -9 87128 10.21010 -0.967 0.336 ## — ## Signif. codes: 0 '***' ' 0.001 '** ' 0.01 0.05 '.' 0.1 ## Residual standard error: 0.5424 on 82 degrees of freedom ## Multiple R-squared: 0.9003, Adjusted R-squared: 0.8942 ## F-statistic: 148.1 on 5 and 82 DF, p-value: < 2.2e-16 DIAGNOSTIC TEST vif(sml) ## blip ebtllp equity npllog gdp ## 1.479442 4.355729 1.806039 4.154619 1.012291 bgtest(sml) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: sml ## LM test = 0.23988, df = 1, p-value = 0.6243 dwtest(sml) ## Durbin-Watson test ## data: sml ## DW = 1.8858, p-value = 0.265 ## alternative hypothesis: true autocorrelation is greater than 0 bptest(sml) ## studentized Breusch-Pagan test ## data: sml ## BP = 1.3718, df = 5, p-value = 0.9274 bptest(sml,varformula=~ fitted.values(sml),studentize = TRUE,data=smoothpl) ## studentized Breusch-Pagan test ## data: sml ## BP = 0.0058901, df = 1, p-value = 0.9388 resettest(sml,type="regressor") ## RESET test ## data: sml ## RESET = 0.90613, dfl = 10, df2 = 72, p-value = 0.5323 resettest(sml,type="fitted") ## RESET test ## data: sml ## RESET = 0.63283, dfl = 2, df2 = 80, p-value = 0.5337 RELAIMPO - SMOOTHING MODEL sml.rela<-calc•relimp.Im(sml) print(sml.rela) ## Response variable: lip ## Total response variance: 2.781423 ## Analysis based on 88 observations ## 5 Regressors: ## blip ebtllp equity npllog gdp ## Proportion of variance explained by model: 90.03% ## Metrics are not normalized (rela=FALSE). ## Relative importance metrics: ## Img ## blip 0.0826840310 ## ebtllp 0.3547947533 ## equity 0.1237212394 ## npllog 0.3381566363 ## gdp 0.0009563398 ## Average coefficients for different model sizes: ## IX 2Xs 3Xs 4Xs 5Xs ## blip 182.782275 91.6287167 42.1740172 20.5752839 12.8812897 ## ebtllp 1.046881 0.9114817 0 7883701 0.6774991 0.5790958 ## equity -15.162719 -8.1736019 -3.8964895 -1.6629153 -0.8053896 ## npllog 1.222225 1.0328839 0.8642960 0.7162487 0.5891239 ## gdp -7.486676 -5.5417713 -7.6124919 -9.3061670 -9 8712783 plot(sml.rela) Relative importances for lip R2 = 90.03%, metrics are not normalized. GIAI ĐOẠN 2012 ĐÉN 2015 sm2 <-lm(llp~bllp+ebtllp+equity+npllog+gdp,data=smoothp2) summary(sm2) ## Call: ## lm(formula = lip ~ blip + ebtllp + equity + npllog + gdp, data=smoothp2) ## Residuals: ## Min IQ Median 3Q Max ## -1.00183 -0.23735 -0.00561 0.26632 0.78851 ## Coefficients: ## Estimate std. Error t value Pr(>|t1) ## (Intercept) -3.34155 1.17027 -2.855 0.00556 ** ## blip 22.79973 6 72380 3.391 0.00111 ** ## ebtllp 0.70348 0.06235 11.283 < 2e-16 *** ## equity -0.67800 1.67384 -0.405 0.68659 ## npllog 0.38074 0.08881 4.287 5.32e-05 *** ## gdp 18.82451 9.08134 2 073 0.04162 * ## — ## Signif. 1 codes: 0 '***' 0.001 ”1 |e*' 0.01 0.05 '.' 0.1 ## Residual standard error: 0.4244 on 75 degrees of freedom ## (7 observations deleted due to missingness) ## Multiple R-squared: 0.9208, Adjusted R-squared: 0.9155 ## F-statistic: 174.3 on 5 and 75 DF, p-value: < 2.2e-16 DIAGNOSTIC TEST vif(sm2) ## blip ebtllp equity npllog gdp ## 1.196060 3.187835 2.058255 4.026566 1.099096 bgtest(sm2) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: sm2 ## LM test = 2.1308, df = 1, p-value = 0.1444 dwtest(sm2) ## Durbin-Watson test ## data: sm2 ## DW = 1.6727, p-value = 0.04678 ## alternative hypothesis: true autocorrelation is greater than 0 bptest(sm2) ## studentized Breusch-Pagan test ## data: sm2 ## BP = 9.8642, df = 5, p-value = 0.07918 bptest(sm2,varformula= ~fitted.values(sm2),studentize = TRUE,data=smoothp2) ## studentized Breusch-Pagan test ## data: sm2 ## BP = 2.4108, df = 1, p-value = 0.1205 resettest(sm2,type="regressor") ## RESET test ## data: sm2 ## RESET = 1.4492, dfl = 10, df2 = 65, p-value = 0.1793 resettest(sm2,type="fitted") ## RESET test ## data: sm2 ## RESET = 1.7067, dfl = 2, df2 = 73, p-value = 0.1886 RELAIMPO - SMOOTHING MODEL sm2.rela<-calc.relimp.Im(sm2) print(sm2.rela) ## ## ## ## ## ## ## ## Response variable: lip Total response variance: 2.13136 Analysis based on 81 observations 5 Regressors: ## blip ebtllp equity npllog Proportion of variance explained by model: Metrics are not normalized (rela=FALSE). Relative importance metrics: Img 0.05648723 0.43721120 0.10068431 0.31013970 0.01625120 coefficients for different model ## blip ## ebtllp ## equity ## npllog ## gdp ## Average ## ## blip ## ebtllp ## equity ## npllog ## gdp IX 66.519582 1.005918 -19.552828 1.167496 58.020702 2Xs 49.8020501 0.9221918 -10.4075487 0.9601426 31.6849927 3Xs 37.0775597 0.8396605 -4.4709145 0 7572983 22.1585977 gdp 92.08% sizes: 4Xs 27.3586909 0.7640478 -1.2811661 0.5639565 20.9694159 5Xs 22.7997274 0.7034810 -0.6780001 0.3807378 18.8245147 plot(sm2.rela) Relative importances for lip R = 92.08%, metrics are not normalized. PHẦN 3 - KẾT QUẢ HỒI QUY DỮ LIỆU BẢNG (PANEL DATA) library(plm) ## Loading required package: Formula psmooth<-plm.data(smooth,indexes=c("bank","year")) MÔ HÌNH TẢCĐỘNG Gộp smpool<-plm(llp~bllp+ebtllp+equity+npllog+gdp,data=psmooth,model="pooling") summary(smpool) ## Oneway (individual) effect Pooling Model ## ## Call: ## plm(formula = lip ~ blip + ebtllp + equity + npllog + gdp, data = psmoot h, ## model = "pooling") ## ## Unbalanced Panel: n=22, T=6-8, N= =169 ## ## Residuals : ## Min. 1st Qu. Median 3rd Qu. Max. ## -2.0300 -0. 3100 0.0796 0.3190 1.2900 ## ## Coefficients : ## Estimate std. Error t-value Pr(>|t|) ## (Intercept) -3.605329 0.789254 -4.5680 9.669e-06 *** ## blip 30.649690 6.669026 4.5958 8.595e-06 *** ## ebtllp 0.557803 0.050690 11.0042 < 2.2e-16 *** ## equity -0.766654 0.911323 -0.8413 0.4014 ## npllog 0.617960 0.057314 10.7821 < 2.2e-16 *** ## gdp -2.665589 7.118830 -0.3744 0.7086 ## — ## Signif. codes: 0 '***' 0.001 '*=• " 0.01 ' '*• 0.05 '. ' 0 ## ## Total Sum of Squares: 502.04 ## Residual Sum of Squares: : 44.923 ## R-Squared: 0.91052 ## Adj. R-Squared: 0.87819 ## F-statistic: 331.728 on 5 and 163 DF, p-value: < 2.22e-16 DIGNOSTIC TEST FOR POOLING MODEL pbgtest(smpool) ## Breusch-Godfrey/Wooldridge test for serial correlation in panel models ## data: lip ~ blip + ebtllp + equity + npllog + gdp ## chisq = 2.0168, df = 6, p-value = 0.9181 ## alternative hypothesis: serial correlation in idiosyncratic errors bptest(smpool) ## studentized Breusch-Pagan test ## data: smpool ## BP = 2.6211, df = 5, p-value = 0.7582 MÔ HÌNH TẢC ĐỘNG CÓ ĐỊNH (FEM) smfe<-plm(llp~bllp+ebtllp+equỉty+npllog+gdp,model="wỉthỉn",data=psmooth) summary(smfe) ## Oneway (individual) effect Within Model ## ## Call: ## plm(formula = lip ~ blip + ebtllp + equity + npllog + gdp, data = psmoot h, ## model = "within") ## ## Unbalanced Panel: n=22, T=6-8, N=169 ## ## Residuals : ## Min. 1st Qu. Median 3rd Qu. Max. ## -1.8000 -0.2810 0.0452 0.3370 1.2300 ## ## Coefficients : ## Estimate std. Error t-value Pr(>|t|) ## blip 14.507128 7.762764 1.8688 0.06371 • ## ebtllp 0.600462 0.075542 7.9487 5.308e-13 *** ## equity -0.121487 1.177842 -0.1031 0.91799 ## npllog 0.730106 0.065368 11.1692 < 2.2e-16 *** ## gdp -2.683100 6.979501 -0.3844 0.70124 ## — ## Signif. codes: 0 '***' 0.001 '**' 0 1.01 0. 05 ## ## Total Sum of Squares: 155.1 ## Residual Sum of Squares: 37.177 ## R-Squared: 0.76031 ## Adj. R-Squared: 0.63884 ## F-statistic: 90.0867 on 5 and 142 DF, p-value: < 2.22e-16 DIGNOSTIC TEST FOR FEM pbgtest(smfe) ## Breusch-Godfrey/Wooldridge test for serial correlation in panel models ## data: lip ~ blip + ebtllp + equity + npllog + gdp ## chisq = 15.559, df = 6, p-value = 0.01633 ## alternative hypothesis: serial correlation in idiosyncratic errors bptest(smfe) ## studentized Breusch-Pagan test ## data: smfe ## BP = 2.6211, df = 5, p-value = 0.7582 MÔ HÌNH TẢC ĐỘNG NGẪU NHIÊN (REM) smre<-plm(llp~bllp+ebtllp+equity+npllog+gdp,model="random",data=psmooth) summary(smre) ## Oneway (individual) effect Random Effect Model ## (Swamy-Arora's transformation) ## ## Call: ## plm(formula = lip ~ blip + ebtllp + equity + npllog + gdp, data = psmoot h, ## model = "random") ## ## Unbalanced Panel: n=22, T=6-8, N=169 ## ## Effects: ## Warning in sqrt(sigma2): NaNs produced ## var std.dev share ## idiosyncratic 0.26181 0.51167 1.045 ## individual -0.01131 NA -0.045 ## theta : ## Min. 1st Qu. Median Mean 3rd Qu. Max. ## -0.2361 -0.2361 -0.2361 -0.2254 -0.2361 -0.1618 ## ## Residuals : ## Min. 1st Qu. Median Mean 3rd Qu. Max. ## -2.06000 -0.32600 0.04770 -0.00115 0.34600 1.35000 ## ## Coefficients : ## Estimate std. Error t-value Pr(>|t|) ## (Intercept) -3.261725 0.750998 -4.3432 2.460e-05 *** ## blip 35.437866 6.410403 5.5282 1.261e-07 *** ## ebtllp 0.561132 0.047875 11.7208 < 2.2e-16 *** ## equity -0.968146 0.856489 -1.1304 0.2600 ## npllog 0.586918 0.055714 10.5344 < 2.2e-16 *** ## gdp -2.507903 7.326281 -0.3423 0.7326 ## — ## Signif. codes: 0 '***' 0.001 '** ' 0.01 1 0.05 '. ' 0 ## ## Total Sum of Squares: 750.57 ## Residual Sum of Squares : 47.762 ## R-Squared: 0.93638 ## Adj. R-Squared: 0.90314 ## F-statistic: 479.703 on 5 and 163 DF, p-value: < 2. 22e DIGNOSTIC TEST FOR REM pbgtest(smre) ## Breusch-Godfrey/Wooldridge test for serial correlation in panel models ## ## data: lip ~ blip + ebtllp + equity + npllog + gdp ## chỉsq = 0.93188, df = 6, p-value = 0.9881 ## alternative hypothesis: serial correlation in idiosyncratic errors bptest(smre) ## studentized Breusch-Pagan test ## data: smre ## BP = 2.6211, df = 5, p-value = 0.7582 CÁC KIỀM ĐỊNH LựA CHỌN MÔ HÌNH TÓI ưu FIXED EFFECT MODEL vs. POOLING MODEL (P value is small =>Fixed effect model) pFtest(smfe,smpool) ## ## F test for individual effects ## ## data: lip ~ blip + ebtllp + equity + npllog + gdp ## F = 1.4088, dfl = 21, df2 = 142, p-value = 0.1234 ## alternative hypothesis: significant effects FIXED EFFECT MODEL vs. RANDOM EFFECT MODEL (p value is small =fixed effect model) phtest(smfe,smre) ## ## Hausman Test ## ## data: lip ~ blip + ebtllp + equity + npllog + gdp ## chisq = 28.981, df = 5, p-value = 2.339e-05 ## alternative hypothesis: one model is inconsistent RANDOM EFFECT vs. POOLING MODEL (pvalue is small => random effect model) plmtest(smpool) ## Lagrange Multiplier Test - (Honda) ## ## data: lip ~ blip + ebtllp + equity + npllog + gdp ## normal = 0.34703, p-value = 0.7286 ## alternative hypothesis: significant effects PHỤ LỤC 8. KẾT QUẢ HỒI QUY MÔ HÌNH DAO ĐỘNG LUỒNG TIỀN VÀ LỢI NHUẬN PHẦN 1. KẾT QUẢ HỒI QUY MÔ HÌNH Độ DAO ĐỘNG LUỒNG TIỀN GIAI ĐOẠN TỪ2008 ĐÉN 2015 cm <-lm(dcash~bcfo+bebt+deposỉt+ovh+loan+growth,data=Var) summary(cm) ## Call: ## lm(formula = deash ~ bcfo + bebt + deposit + ovh + loan + growth, ## data = Van) ## Residuals: ## Min IQ Median 3Q Max ## -0.28380 -0.05094 -0.00319 0.04432 0.39987 ## Estimate std. Error t value Pr(>|t|) ## (Intercept) 0.04848 0.04049 1.197 0.2329 ## bcfo -1.34877 0 07370 -18.301 <2e-16 *** ## bebt 2.34427 1.01574 2.308 0.0222 * ## deposit 0.06851 0.06020 1.138 0.2568 ## ovh -3.19356 1.59695 -2.000 0.0472 * ## loan -0.13661 0.05934 -2 302 0.0226 * ## growth 0.29349 0.03009 9.753 <2e-16 *** ## — ## Signif. codes: 0 '***' 0.001 ■ =1 |e*' 0.01 0.05 0.1 ## Coefficients: ## Residual standard error: 0.09781 on 166 degrees of freedom ## (7 observations deleted due to missingness) ## Multiple R-squared: 0.7382, Adjusted R-squared: 0.7287 ## F-statistic: 78.01 on 6 and 166 DF, p-value: < 2.2e-16 DIAGNOSTIC TEST FOR CM vif(cm) ## bcfo bebt deposit ovh loan growth ## 1.073735 1.163129 2.318072 1.465930 2.447562 1.706712 dwtest(cm) ## Durbin-Watson test ## data: cm ## DW = 2.2049, p-value = 0.8835 ## alternative hypothesis: true autocorrelation is greater than 0 bgtest(cm) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: cm ## LM test = 3.3657, df = 1, p-value = 0.06657 bptest(cm) ## studentized Breusch-Pagan test ## data: cm ## BP = 28 855, df = 6, p-value = 6.482e-05 bptest(cm,varformula = ~fitted.values(cm),studentize = TRUE,data=Var) ## studentized Breusch-Pagan test ## data: cm ## BP = 0.081549, df = 1, p-value = 0.7752 resettest(cm,type="fitted") ## RESET test ## data: cm ## RESET = 0.47028, dfl = 2, df2 = 164, p-value = 0.6257 RELAIMPO -CASH MODEL cm.rela<-calc.relimp(cm) print(cm.rela) ## Response variable: deash ## Total response variance: 0.03526867 ## Analysis based on 173 observations ## ## 6 Regressors: ## befo bebt deposit ovh loan growth ## Proportion of variance explained by model: 73.82% ## Metrics are not normalized (rela=FALSE). ## ## Relative importance metrics: ## ## Img ## befo 0.507965284 ## bebt 0.006118172 ## deposit 0.033002360 ## ovh 0.012081594 ## loan 0.018830116 ## growth 0.160207631 ## ## Average coefficients i for different model sizes: ## ## IX 2Xs 3Xs 4Xs 5Xs ## befo -1.2793084 - 1.2650683 -1.27818256 -1.30231132 -1.32653171 ## bebt 1 6587207 1.1916237 1.26576921 1.57244563 1.95185077 ## deposit 0.3029049 0.2254084 0.16742389 0.12390690 0.09168498 ## ovh 6.4974477 3.2092020 0.81471859 -0 86982293 -2.09842770 ## loan 0.2324043 0.1239518 0.03200797 -0.04230544 -0.09845039 ## growth 0.2591282 0.2540758 0.26191471 0.27439978 0.28537507 ## 6Xs ## befo -1.3487735 ## bebt 2.3442719 ## deposit 0.0685081 ## ovh -3.1935612 ## loan -0.1366144 ## growth 0.2934852 plot(cm.rela) Relative importances for deash Method LMG R2 = 73 82%, metrics are not normalized. GIAI ĐOẠN TỪ2008 ĐÉN 2011 cml <-lm(dcash~bcfo+bebt+deposỉt+ovh+loan+growth,data=Varl) summary(cml) ## Call: ## lm(formula = deash ~ befo + bebt + deposit + ovh + loan ■ ## data = Vari) ## Residuals: ## Min IQ Median 3Q Max ## -0.26865 -0. 06099 0, .00064 0.04747 0.36411 ## ## Coefficients I ## Estimate std. Error t value Pr(>|t|) ## (Intercept) 0.08324 0.06422 1.296 0.19870 ## befo -1.46940 0.12476 - 11 778 < 2e-16 *** ## bebt 1.71842 1.56628 1.097 0.27592 ## deposit 0.21094 0.09981 2.113 0 03772 * ## ovh -2 31727 3.05447 -0.759 0.45032 ## loan -0.30627 0.10551 -2.903 0.00479 ** ## growth 0.26598 0.04502 5.907 8.34e-08 *** ## — ## Signif. codes: 0 '***' 0.001 '** ' 0.01 0.05 '.' 0.1 ## ## Residual standard error: 0.1127 on 79 degrees of freedom ## (2 observations deleted due to missingness) ## Multiple R-squared: 0.7286, Adjusted R-squared: 0.708 ## F-statistic: 35.35 on 6 and 79 DF, p-value: < 2.2e-16 DIAGNOSTIC TEST FOR CM1 vif(cml) ## bcfo bebt deposit ovh loan growth ## 1.129039 1.030178 2.887863 2.098580 3.396329 1.825758 dwtest(cml) ## Durbin-Watson test ## data: cml ## DW = 2.1265, p-value = 0.6735 ## alternative hypothesis: true autocorrelation is greater than 0 bgtest(cml) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: cml ## LM test = 0.92726, df = 1, p-value = 0.3356 bptest(cml) ## studentized Breusch-Pagan test ## data: cml ## BP = 15.415, df = 6, p-value = 0.01727 bptest(cml,varformula = ~ fitted.values(cml),studentize = TRUE,dat=Varl) ## studentized Breusch-Pagan test ## data: cml ## BP = 0.12382, df = 1, p-value = 0.7249 resettest(cml,type="regressor") ## RESET test ## data: cml ## RESET = 1.2939, dfl = 12, df2 = 67, p-value = 0.243 resettest(cml,type="fitted") ## RESET test ## data: cml ## RESET = 1.7219, dfl = 2, df2 = 77, p-value = 0.1855 RELAIMPO -CASH MODEL cml.rela<-calc.relimp(cml) print(cml.rela) ## Response variable: deash ## Total response variance: 0.04351478 ## Analysis based on 86 observations ## ## 6 Regressors: ## befo bebt deposit ovh loan growth ## Proportion of variance explained by model: 72.86% ## Metrics are not normalized (rela=FALSE). ## ## Relative importance metrics: ## ## Img ## bcfo 0.461691047 ## bebt 0.003448727 ## deposit 0.059776290 ## ovh 0.015978261 ## loan 0 028352298 ## growth 0.159382942 ## ## Average coefficients for different model sizes: ## ## IX 2Xs 3Xs 4Xs 5Xs ## bcfo -1.4060845 -1.36283795 -1.37925057 -1.4152130 -1.4472569 ## bebt 1.5551312 1.34492683 1.39958063 1.5312639 1.6408298 ## deposit 0.3756777 0 29737628 0.25753080 0.2392347 0.2278543 ## ovh 8.6249729 3 68892391 0.65413202 -0.9953740 -1.7976917 ## loan 0.2505038 0.08735735 -0.05643429 -0.1740177 -0.2592030 ## growth 0.2757853 0.25908784 0.26106198 0 2667682 0.2685176 ## 6Xs ## bcfo -1.4694027 ## bebt 1.7184191 ## deposit 0.2109442 ## ovh -2.3172735 ## loan -0.3062652 ## growth 0.2659787 plot(cml.rela) Relative importances for deash Method LMG R2 = 72.86%, metrics are not normalized. GIAI ĐOẠN TỪ2012 ĐÉN 2015 cm2 <-lm(dcash~bcfo+bebt+deposỉt+ovh+loan+growth,data=Var2) summary(cm2) ## Call: ## lm(formula = deash ~ befo + bebt + deposit + ovh + loan + growth, ## data = Var2) ## ## Residuals: ## Min IQ Median 3Q Max ## -0.196897 -0.034964 0.001906 0.032627 0.240967 ## ## Coefficients: ## Estimate std. Error t value Pr(>|t1) ## (Intercept) -0.0169128 0.0546430 -0.310 0.75773 ## befo -1.3233471 0.0837980 -15.792 < 2e-16 *** ## bebt 0.5490166 1.3645909 0.402 0.68851 ## deposit 0.0008785 0 0876398 0.010 0.99203 ## ovh -1.0095604 1.7253180 -0.585 0.56010 ## loan 0.0135551 0.0758542 0.179 0.85863 ## growth 0.2179600 0.0685830 3 178 0.00211 ** ## — ## Signif. codes: 0 '***■ 0.001 '**' ' 0.01 '* ' 0.05 '.' 0.1 ## ## Residual standard error: 0.0768 on 80 degrees of freedom ## (1 observation deleted due to missingness) ## Multiple R-squared: 0.7922, Adjusted R-squared: 0.7766 ## F-statistic: 50.84 on 6 and 80 DF, p-value: < 2.2e-16 DIAGNOSTIC TEST FOR CM2 vif(cm2) ## bcfo bebt deposit ovh loan growth ## 1.054774 1.357913 3.171440 1.332476 2.271640 2.287288 dwtest(cm2) ## Durbin-Watson test ## data: cm2 ## DW = 2.0231, p-value = 0.4564 ## alternative hypothesis: true autocorrelation is greater than 0 bgtest(cm2) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: cm2 ## LM test = 0.022622, df = 1, p-value = 0.8804 bptest(cm2) ## studentized Breusch-Pagan test ## data: cm2 ## BP = 21.42, df = 6, p-value = 0.001541 bptest(cm2,varformula = ~fitted.values(cm2),studentize = TRUE,data=Var2) ## studentized Breusch-Pagan test ## data: cm2 ## BP = 0.31179, df = 1, p-value = 0.5766 resettest(cm2,type="fitted") ## RESET test ## data: cm2 ## RESET = 3.3037, dfl = 2, df2 = 78, p-value = 0.04195 RELAIMPO -CASH MODEL cm2.rela<-calc.relimp(cm2) print(cm2.rela) ## Response variable: deash ## Total response variance: 0.02640569 ## Analysis based on 87 observations ## 6 Regressors: befo bebt deposit ovh loan growth ## Proportion of variance explained by model: 79.22% ## Metrics are not normalized (rela=FALSE). ## ## Relative importance metrics: ## Img ## befo 0.688176783 ## bebt 0.001034278 ## deposit 0.015729134 ## ovh 0.006943530 ## loan 0.008170101 ## growth 0.072174157 ## Average coefficients for different model sizes: ## IX 2Xs 3Xs 4Xs 5Xs ## befo -1.3697219 -1.35515608 -1.34502535 -1.33728906 -1.330345005 ## bebt -0.2188242 -0.64108499 -0.52002498 -0.23550414 0.113361544 ## deposit 0.1975642 0.11780990 0.05955271 0.01965427 -0.000607236 ## ovh 4.3296975 2.75661241 1.59857323 0.65911208 -0.176162072 ## loan 0.1576639 0.08246534 0.04018371 0.02294709 0.017599192 ## growth 0.3198158 0.32106451 0 30867671 0 28578358 0.254913711 ## 6Xs ## befo -1.3233471197 ## bebt 0.5490165779 ## deposit 0.0008785082 ## ovh -1.0095604122 ## loan 0.0135550939 ## growth 0.2179600420 plot(cm2.rela) Relative importances for deash Method LMG <D Is- CD _ > o _ CO XT £Z o — Ó o'-- o —I o -1 befo bebt depo ovh loan grow R2 = 79.22%, metrics are not normalized. PHẦN 2: KẾT QUẢ HỒI QUY MÔ HÌNH Độ DAO ĐỘNG CỦA LỢI NHUẬN GIAI ĐOẠN 2008 ĐÉN 2015 dm <-lm(dnỉ~bcfo+bebt+deposỉt+ovh+loan+growth,data=Var) summary(dm) ## Call: ## lm(formula = dnỉ ~ befo + bebt + deposit + ovh + loan + growth, ## data = Var) ## Residuals: ## Min IQ Median 3Q Max ## -0.037239 -0.004074 -0.000334 0.003532 0.050726 ## Coefficients • ## Estimate std. Error t value Pr(>|t|) ## (Intercept) -0.0004358 0.0033795 -0.129 0.8975 ## bcfo 0.0140365 0.0061507 2 282 0.0238 * ## bebt -0.5240651 0.0847688 -6.182 4.73e-09 *** ## deposit -0.0017977 0.0050240 -0.358 0.7209 ## ovh 0.0575243 0.1332743 0.432 0.6666 ## loan 0.0087164 0.0049523 1.760 0.0802 • ## growth 0.0142590 0.0025114 5.678 5.97e-08 *** ## — ## Signif. codes: 0 '***' 0.001 '**' 0.01 •• 0.05 '. ' 0.1 ## Residual standard error: 0.008163 on 166 degrees of freedom ## (7 observations deleted due to missingness) ## Multiple R-squared: 0.3884, Adjusted R-squared: 0.3663 ## F-statistic: 17.57 on 6 and 166 DF, p-value: 1.06e-15 DIAGNOSTIC TEST FOR DM vif(dm) ## bcfo bebt deposit ovh loan growth ## 1.073735 1.163129 2.318072 1.465930 2.447562 1.706712 dwtest(dm) ## Durbin-Watson test ## data: dm ## DW = 1.9481, p-value = 0.3062 ## alternative hypothesis: true autocorrelation is greater than 0 bgtest(dm) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: dm ## LM test = 0.124, df = 1, p-value = 0.7247 bptest(dm) ## studentized Breusch-Pagan test ## data: dm ## BP = 22.014, df = 6, p-value = 0.001204 bptest(dm,varformula = ~fitted.values(dm),studentize = TRUE,data=Var) ## studentized Breusch-Pagan test ## data: dm ## BP = 0.20771, df = 1, p-value = 0.6486 resettest(dm,type="fitted") ## RESET test ## data: dm ## RESET = 1.7685, dfl = 2, df2 = 164, p-value = 0.1738 RELAIMPO - DNI MODEL dm.rela<-calc.relimp(dm) print(dm.rela) ## Response variable: dni ## Total response variance: 0.0001051424 ## Analysis based on 173 observations ## 6 Regressors: ## bcfo bebt deposit ovh loan growth ## Proportion of variance explained by model: 38.84% ## Metrics are not normalized (rela=FALSE). ## Relative importance metrics: ## Img ## bcfo 0.016152502 ## bebt 0.116680457 ## deposit 0.045590567 ## ovh 0.006176212 ## loan 0.046025761 ## growth 0.157755202 ## Average coefficients for different model sizes: ## IX 2Xs 3Xs 4Xs 5Xs ## bcfo 0.009323846 0.01174139 0.01280486 0.013228608 0.013611342 ## bebt -0.368090575 -0.42074281 -0.45238500 -0.475562455 -0.499000415 ## deposit 0.019257707 0.01573583 0.01167321 0.007259786 0.002706360 ## ovh 0.217095213 0.08150483 0.01587605 0.003709704 0.024922226 ## loan 0.016790595 0.01363617 0.01122839 0.009606026 0.008780984 ## growth 0.015035155 0.01473704 0.01438976 0.014147407 0.014107174 ## 6Xs ## bcfo 0.014036467 ## bebt -0.524065120 ## deposit -0.001797659 ## ovh 0.057524277 ## loan 0.008716354 ## growth 0.014258984 plot(dm.rela) Relative importances for dni Method LMG bcfo bebt depo ovh loan grow R2 = 38.84%, metrics are not normalized. GIAI ĐOẠN TỪ2008 ĐẾN 2011 dml <-lm(dnỉ~bcfo+bebt+deposỉt+ovh+loan+growth,data=Varl) summary(dml) ## Call: ## lm(formula = dnỉ ~ bcfo + bebt + deposit + ovh + loan + growth, ## data = Vari) ## Residuals: ## Min IQ Median 3Q Max ## -0.023302 -0 .005183 -0. 000257 0.004120 0 .049530 ## Coefficients I ## Estimate std. Error t value Pr(>|t|) ## (Intercept) -0.0004042 0.0054895 -0.074 0.9415 ## bcfo 0.0038581 0.0106643 0.362 0.7185 ## bebt -0.7487216 0.1338790 -5.593 3.1e-07 *** ## deposit 0.0180929 0.0085317 2.121 0.0371 * ## ovh 0.2346814 0.2610838 0.899 0.3715 ## loan -0.0039346 0.0090187 -0.436 0.6638 ## growth 0.0070256 0.0038485 1.826 0.0717 . ## — ## Signif. codes: 0 '***' 0.001 '** ' 0.01 '* ' 0.05 '.' 0.1 ## ## Residual standard error: 0.009635 on 79 degrees of freedom ## (2 observations deleted due to missingness) ## Multiple R-squared: 0.4479, Adjusted R-squared: 0.406 ## F-statistic: 10.68 on 6 and 79 DF, p-value: 1.153e-08 DIAGNOSTIC TEST FOR DM1 vif(dml) ## bcfo bebt deposit ovh loan growth ## 1.129039 1.030178 2.887863 2.098580 3.396329 1.825758 dwtest(dml) ## Durbin-Watson test ## data: dml ## DW = 1.8756, p-value = 0.2028 ## alternative hypothesis: true autocorrelation is greater than 0 bgtest(dml) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: dml ## LM test = 0.36488, df = 1, p-value = 0.5458 bptest(dml) ## studentized Breusch-Pagan test ## data: dml ## BP = 8.5807, df = 6, p-value = 0.1986 bptest(dml,varformula = ~ fitted.values(dml),studentize = TRUE,dat=Varl) ## studentized Breusch-Pagan test ## data: dml ## BP = 1.5035, df = 1, p-value = 0.2201 resettest(dml,type="fitted") ## RESET test ## data: dml ## RESET = 2.7944, dfl = 2, df2 = 77, p-value = 0.06736 RELAIMPO - DNI MODEL dml.rela<-calc.relimp(dml) print(dml.rela) ## Response variable: dni ## Total response variance: 0.0001562766 ## Analysis based on 86 observations ## 6 Regressors: ## bcfo bebt deposit ovh loan growth ## Proportion of variance explained by model: 44.79% ## Metrics are not normalized (rela=FALSE). ## Relative importance metrics: ## Img ## bcfo 0.001247805 ## bebt 0.223519860 ## deposit 0.095340453 ## ovh 0.016106864 ## loan 0.032168206 ## growth 0.079549075 ## Average coefficients for different model sizes • ## IX 2Xs 3Xs 4Xs 5Xs ## bcfo -0.00416618 0.002197624 0.004075882 0.004188767 0.0039992419 ## bebt -0.74948768 ■ -0.757906494 -0.754148258 -0.749367490 -0.7476872061 ## deposit 0.02697414 0.025368076 0.023331820 0.021248396 0.0194264926 ## ovh 0.46342625 0.238121388 0.142060088 0.132281327 0 1728233799 ## loan 0.01863300 0.012940151 0.007803130 0.003286959 -0.0006025325 ## growth 0.01426621 0.012196654 0.010415652 0.009041588 0.0079842902 ## 6Xs ## bcfo 0.003858122 ## bebt -0.748721551 ## deposit 0.018092875 ## ovh 0.234681350 ## loan -0.003934595 ## growth 0.007025649 plot(dml.rela) Relative importances for dni Method LMG 2 J J . • ÉI 1- I R = 44.79%, metrics are not normalized. GIAI ĐOẠN TỪ2012 ĐÉN 2015 dm2 <-lm(dni~bcfo+bebt+deposit+ovh+loan+growth,data=Var2) summary(dm2) ## Call: ## lm(formula = dni ~ bcfo + bebt + deposit + ovh + loan + growth, ## data = Var2) ## Residuals: ## Min IQ Median 3Q Max ## -0.0101489 - 0.0026426 0.0004042 0.0028578 0.0097668 ## ## Coefficients I ## Estimate std. Error t value Pr(>|t|) ## (Intercept) -0.004424 0.003103 -1.426 0.1578 ## bcfo 0.011062 0.004758 2 325 0.0226 * ## bebt -0.442235 0.077483 -5.708 1.87e-07 *** ## deposit 0.004035 0.004976 0.811 0.4198 ## ovh 0.121284 0.097965 1.238 0.2193 ## loan 0.003448 0.004307 0.800 0.4258 ## growth 0.001062 0.003894 0.273 0.7858 ## — ## Signif. codes: 0 ’*** ’ 0.001 ’** ■ 0.01 ■ *■ 0.05 ■.■ 0. ## ## Residual standard error: 0.004361 on 80 degrees of freedom ## (1 observation deleted due to missingness) ## Multiple R-squared: 0.3743, Adjusted R-squared: 0.3274 ## F-statistic: 7.977 on 6 and 80 DF, p-value: 9.355e-07 DIAGNOSTIC TEST FOR DM2 vif(dm2) ## bcfo bebt deposit ovh loan growth ## 1.054774 1.357913 3.171440 1.332476 2.271640 2.287288 dwtest(dm2) ## Durbin-Watson test ## data: dm2 ## DW = 1.9815, p-value = 0.3811 ## alternative hypothesis: true autocorrelation is greater than 0 bgtest(dm2) ## Breusch-Godfrey test for serial correlation of order up to 1 ## data: dm2 ## LM test = 0.0019784, df = 1, p-value = 0.9645 bptest(dm2) ## studentized Breusch-Pagan test ## data: dm2 ## BP = 14.555, df = 6, p-value = 0.02401 bptest(dm2,varformula = ~fitted.values(dm2),studentize = TRUE,data=Var2) ## studentized Breusch-Pagan test ## data: dm2 ## BP = 0.11774, df = 1, p-value = 0.7315 resettest(dm2,type="fitted") ## RESET test ## data: dm2 ## RESET = 1.0798, dfl = 2, df2 = 78, p-value = 0.3447 RELAIMPO - DNI MODEL dm2.rela<-calc.relimp(dm2) print(dm2.rela) ## Response variable: dni ## Total response variance: 2.827054e-05 ## Analysis based on 87 observations ## 6 Regressors: ## bcfo bebt deposit ovh loan growth ## Proportion of variance explained by model: 37.43% ## Metrics are not normalized (rela=FALSE). ## Relative importance metrics: ## Img ## bcfo 0.033958038 ## bebt 0.263129347 ## deposit 0.042343756 ## ovh 0.009301644 ## loan 0.012213302 ## growth 0.013368557 ## ## Average coefficients for different model sizes: ## IX 2Xs 3Xs 4Xs 5Xs ## bcfo 0.008342794 0.008947652 0.009519446 0.009964141 0.010425531 ## bebt -0.376981351 -0.399748606 -0.412872266 -0.421865305 -0.431011794 ## deposit 0.008311812 0.009593354 0.009632617 0.008638796 0.006742146 ## ovh -0.028745835 -0.020818375 0.001930175 0.036131458 0.077298489 ## loan 0.002986003 0.001819711 0.001276151 0.001394335 0.002155694 ## growth 0.004867859 0.004023384 0.002848608 0.001752998 0.001081956 ## 6Xs ## bcfo 0.011062097 ## bebt -0.442235269 ## deposit 0.004035417 ## ovh 0.121283982 ## loan 0.003447771 ## growth 0.001061966 plot(dm2.rela) Relative importances for dni Method LMG 2 m jrtA/ i - i 1- I R = 37.43%, metrics are not normalized.