For Informal Credit Access Model


of household head, distance, interest rate, loan procedures, bank experience, e-banking services with 21 scales and these scales are all among the factors that meet the requirements, have the ability to converge, and represent well the observed variables. The dependent variable scale Household access to formal credit (4 observed variables) has converged and represented well the scales. Thus, through the EFA exploratory factor analysis, it shows that the independent and dependent variables all have the ability to converge and represent well the observed variables in the scale and are included in the next test with CFA analysis.

The results of the correlation coefficient matrix analysis of the variables show that the relationships between the variables included in the analysis are statistically significant. The observed significance levels of the factors: Interest rates, procedures, experience, banking experience, and household head characteristics in the correlation matrix are mostly < 0.01, which shows that these impacts are relatively significant. At the same time, the correlation coefficient r runs from 0.3 < r < 0.7, proving that the variables have an impact on each other and have practical significance.

At the same time, when considering the relationship between the independent variables KC, LS, TTV, KNCH, KNNH, NHĐT, DDCH with the dependent variable Y, it shows that the variables KC, TTV, LS and KNNH have a correlation coefficient of r < 0, which means a negative relationship with the dependent variable. All the remaining independent variables have a correlation coefficient of 0.3 < r < 0.7. Through the analysis, we can see that the correlation coefficient between the independent variables and the dependent variable shows a fairly close correlation with each other. From there, we can put the variables into the CFA model for analysis.

3.3.1.3. Confirmatory factor analysis CFA

After linking the errors to improve the model to fit the actual data, the result of the confirmatory factor analysis CFA has a coefficient of Chi-square/df = 1.889 (< 3); GFI

= 0.885; TLI = 0.926 (> 0.9); CFI=0.94 (> 0.9); RMSEA=0.058 (< 0.08).



Figure 3.1. Results of confirmatory factor analysis CFA

Source: Author's synthesis and analysis

The linking of errors is used to correct the difference between the proposed model and the estimated model. When linking errors, the model will be improved to improve Chi-square. Chi-square is used to measure the level of conformity in more detail of the entire research model with reality. In the model, the smaller the Chi-square, the better. Some authors recommend 1 < χ2/df < 3 (Hair et al., 2016). If the errors are linked together, the covariance between them will decrease and make the Chi-square decrease by a corresponding amount compared to the Chi-square of the original model. Then GFI, TLI, CFI... will also be improved.


The author continues to remove each factor that does not fit the model by examining the standardized Beta coefficients of the indicators in the model. Any Beta coefficient < 0.5 will be eliminated.

All standardized Beta coefficients of the variables are > 0.5 so we can temporarily accept this CFA model.

The results of the scale reliability assessment, exploratory factor analysis EFA and confirmatory factor analysis CFA with the measurement criteria of the hypotheses remaining the same from the beginning of the study are as follows:

H1: There is a positive relationship between collateral value and households' ability to access formal credit.

H2: There is a positive relationship between income and household access to formal credit.

H3: There is a positive relationship between the number of years of business experience and the household's ability to access formal credit.

H4: There is a negative relationship between geographical distance and households' access to formal credit.

H5: There is an inverse relationship between loan interest rates and households' access to formal credit.

H6: There is an inverse relationship between the complexity of loan procedures and households' ability to access formal credit.

H7: There is a positive relationship between bank experience and households' access to formal credit.

H8: There is a positive relationship between the quality of e-banking services and households' access to formal credit.

3.3.1.4. SEM structural model factor analysis

The coefficients in the model fit the actual data, the results have coefficient Chi-square/df = 1.742 (< 3); GFI = 0.896; TLI = 0.938 (> 0.9); CFI=0.950 (> 0.9);

RMSEA=0.053 (< 0.08). The model shows the influence of factors affecting the credit accessibility of commercial banks including 8 factors mentioned by the author before.


The results of the SEM structural analysis demonstrated that the factors: household head characteristics (including collateral and income), household head experience, distance, loan procedures, interest rates, banking experience and e-banking services have an impact on the household's ability to access bank credit. The Beta coefficients all satisfy the initial hypothesis except for the factor of banking experience. The results obtained from reality are contrary to the initial assumption. The more banking experience, the lower the household's ability to access credit.


Figure 3.2. SEM structural model analysis results

(Source: Author's synthesis and analysis)

The above results answered the research question about the direction of impact of the factors. The results confirmed that the independent variables have an impact on the ability of households to access formal credit. However, in this study, the experience of the bank has an impact in the opposite direction compared to the initial hypothesis.


3.3.2. For informal credit access models

3.3.2.1. Assessment of scale reliability

Data collected from the survey was analyzed and evaluated for Cronbach's Alpha coefficient. The results of the reliability assessment of the informal credit access scale are summarized in the table in the appendix. (Detailed analysis results are in the appendix)

The test results show that the Cronbach's Alpha coefficient of all scales is above 0.7, so these scales ensure reliability. The total correlation coefficient of the observed variables is greater than 0.3, which is considered satisfactory. However, the observed variable HQ4: "Consumer credit will help me have the opportunity to increase my income or solve my needs" has a Cronbach's Alpha coefficient if the variable is eliminated, which is greater than the general Cronbach's Alpha coefficient (0.721 > 0.707), so this variable is eliminated to increase reliability for the next EFA analysis.

3.3.2.2. Results of exploratory factor analysis EFA

After running EFA for the first time, the observed variables TL4, TL5 were eliminated because they had loading factors less than 0.5 and the observed variables DK4 and TL6 were also eliminated because they did not converge with the observed variables corresponding to the independent variables. After running EFA for the second time, the variables DK1, DK2, DK3 were eliminated because they had loading factors less than 0.5.

Table 3.9. KMO and Bartlett test


KMO

0.894

Bartlett test

5675.71

Sig.

0.000

Maybe you are interested!

Source: Author compiled from data analysis

The results of exploratory factor analysis with observed variables of households that have not used black credit have good results. First, the KMO coefficient = 0.894 > 0.5 shows that factor analysis is suitable for the research data. Next, the Bartlett test is 5675.71 with a significance level of Sig. = 0.000 < 0.05, meaning that the hypothesis that the observed variables are not correlated with each other in the population can be rejected. Therefore, the hypothesis that the factor model is inappropriate will be rejected, which proves that the data used for analysis is completely suitable. The value of the total variance extracted for the 6th factor is 64.9% > 50% and the convergence coefficient of eigenvalues ​​of this factor is 1.2 > 1, showing that the observed variables begin to converge at 6 factors, these factors explain 64.9% of the variation in the survey data. Therefore, the factors ensure the ability to represent the original survey data.


Similarly, for the dependent variable, with the KMO test coefficient = 0.758, Sig = 0.000, the extracted variance reached 90.43%, showing the ability to converge and represent well the observed variables in the scale.

Thus, after testing the reliability and value, the scales that do not meet the needs are eliminated and the remaining selected scales have been tested to ensure the requirements. Thus, the research model will be adjusted as follows:

3.3.2.3. Confirmatory factor analysis CFA

Table 3.10. Summary of the first CFA analysis results



Scale

Composite Reliability

Average Variance Extracted


The indicators

Social impact

0.831

0.482

Square/df=2.435 <3

Security

0.88

0.647

GFI=0.905>0.9

CFI=0.942>0.9

Intended use

0.948

0.858

TLI=0.932>0.9

Expected Effort

0.786

0.486

RMSEA=0.051

Expected performance

0.745

0.424

Financial literacy

0.792

0.562


Convenient

0.581

0.322


Source: Author compiled from data analysis

After assessing the reliability of the scale and analyzing the EFA exploratory factor, the factors on the intention to use black credit all have high convergence, well representing the observed variables, the next step will continue to analyze the CFA confirmatory factor to confirm the value, reliability and discrimination of the scale. The author uses AMOS 20 software to analyze the CFA confirmatory factor. To see the analysis results clearly, the author has synthesized the results of the first CFA analysis for the scale.

Looking at the table above, we see that some AVEs of the independent variables are less than 0.5, so we need to remove some observed variables to improve this index. Specifically, after analyzing and re-running CFA, the author concluded to remove 2 independent factor variables, expected performance and convenience, from the scale. Along with that are 2 observed variables, NL1 and AH5, to improve the total variance extracted.


The results of the second CFA analysis of the model's suitability indicators show that the Chi-quare/df value = 2.201 < 3, TLI = 0.967, CFI = 0.974, GFI = 0.948 are all greater than 0.9, the RMSEA coefficient = 0.046 <0.05, so the model is suitable for the market. In addition, all AVE values ​​> 0.5, the value of the composite reliability > 0.7, the reliability of the scale is guaranteed, the discrimination is guaranteed AVE > MSV. The P-value coefficient of the observed variables representing the factors is < 0.5, these observed variables are capable of representing the factors well in the CFA model. From there, it can be affirmed that the scale achieves convergent value and unidimensionality. Thus, the research scales for individual business households accessing informal credit have ensured the analytical requirements.

Table 3.11. Summary of the results of the second CFA analysis



Scale


Composite Reliability

Average Variance

Extracted )


MSV


The indicators

Social impact

0.831

0.524

0.441

Chisquare/df=2.201





<3

Security

0.88

0.647

0.135

GFI=0.948>0.9

Intended use

0.948

0.858

0.441

CFI=0.974>0.9

Expected Effort

0.786

0.524

0.25

TLI=0.967>0.9

Financial literacy

0.792

0.562

0.242

RMSEA=0.046<0.5

Source: Author compiled from data analysis

The correlation coefficients between the components and the standard deviations of the scales are all different from 1 at the 95% confidence level, reaching statistical significance (all P-values ​​are 0). Financial literacy and Expected effort both have discriminant values ​​with correlations between the components of the scale.


Table 3.12. Testing the correlation of variables in the informal credit access model


Correlate

Estimate

(Estimate)


SE


CR

P-

value

SOCIETY

<-->

BAOMAT

.430

.044

9,772

***

SOCIETY

<-->

YDINH

.522

.047

11,201

***

SOCIETY

<-->

NOLUC

.285

.034

8,271

***

SOCIETY

<-->

HIEUBIET

.269

.039

6,928

***

BAOMAT

<-->

YDINH

.336

.045

7,442

***

BAOMAT

<-->

NOLUC

.237

.036

6.632

***

BAOMAT

<-->

HIEUBIET

.261

.042

6,197

***

YDINH

<-->

NOLUC

.184

.036

5.167

***

YDINH

<-->

HIEUBIET

.257

.043

5,963

***

NOLUC

<-->

HIEUBIET

.308

.037

8,324

***

Source: Author's synthesis and data analysis

The results of the correlation test of the components of the intention to use scale above show that after CFA analysis, the household intention to use informal credit scale consists of 4 independent variables (Expected Effort, Social Influence, Security and Financial Knowledge) with 18 observed variables and a dependent variable (Intention to Use) with 3 observed variables. The CFA results show that the components of the scale all achieve structural validity including: convergent validity, discriminant validity, similarity validity with correlation and meet the validity requirements.

3.3.2.4. SEM structural model factor analysis

Testing the relationship between factors in the structural model

The relationship model between factors in the structural model of informal credit access of individual business households is shown in Table 3.12.

Comment


Agree Privacy Policy *