The relationship between financial behavior and investment performance in the real estate market of Ho Chi Minh city

showed that investors' overconfidence did reduce investment performance; Nyamute et al (2015), Muriithi (2016) studied the relationship between investor behavior and portfolio efficiency on the Nairobi Stock Exchange found that overconfidence of investors affects targets. investment portfolio efficiency; Vuong Duc Hoang Quan and Bui Chien Cong (2016) with research on factors influencing decisions of Vietnamese securities investors show that individual investors are influenced by overconfidence factors. and the psychology of loss making investment performance higher. This shows that self-confident individual investors increase their ability to control investment decisions, make investors more confident, more decisive, contributing to bring profits, better efficiency for investors; Obong’o (2016) believes that behavioral factors such as overconfidence play an important and positive role in investment performance as well as for the development of the real estate market. The author gives a third hypothesis as follows: H3: Overconfidence has a positive relationship with investment performance. 2.4 Proposed research model 10 CHAPTER 3: RESEARCH DESIGN 3.1 Research Process 3.1.1 Preliminary research 3.1.1.1. Qualitative research Step 1: Identify research problems (research reasons), review relevant domestic and foreign studies, ask research questions, research objectives, subjects and scope, and identify basis theory, empirical studies, from which build a preliminary scale (preliminary questionnaire). Step 2: Interview experts, by group discussion. Some contents related to group discussion are shown as follows: + Who is an expert: Up to now, there are no studies to identify experts for which standards are outlined for experts. However, the dissertation has also selected different experts in many fields that are directly and indirectly related to the research objects in the thesis. + Number of experts: Up to now, there are no studies to show which standards are outlined for the number of experts. However, the average number of experts is approximately 9 to 10 people, and the number of experts in this study is 17 experts, divided into 2 groups, the first and the second group have the number of 8 respectively. and 9 members. + Purpose of discussion: For the first group (including representatives of state management agencies, lecturers of universities, banks, real estate enterprises, real estate brokers) to discuss about the Psychological or behavioral factors of individual investors affect investment efficiency and psychological factors of investors are affected by external environmental factors, thereby determining research model. This is also a target for discussion of the first group; On the basis of identified the proposed research model from the first group discussion, then discuss with the second group (mainly individual investors directly investing in the HCMC real estate market) about exploring the scale of the concepts of investment efficiency, psychological factors, external environmental factors that indirectly affect the investment efficiency of investors in the real estate market. Since then, it is determined to keep, adjust and supplement the observed variables for the scales in accordance with the space and time at the time of the study, this is also the goal of discussion of the second group. 3.1.1.2. Preliminary quantitative research The survey questionnaire is completed through qualitative research, the author issues the survey directly to the survey object and processes the data through the SPSS 20 software, thereby assessing the reliability. using the scales with Cronbach's Alpha reliability coefficients and exploratory analysis, this is the goal of the preliminary research step before conducting the survey for official quantitative research. In this content, the research of the thesis is conducted sequentially through the following two steps: Step 1: According to Creswell (2003), a targeted survey is a purposeful non-probability sampling, where people are selected because they know best about the survey topic. The author has collected one-time research data with a small number of samples with the aim of evaluating the reliability coefficient of the scales and analyzing the discovery factor to find the distinction of the factors, with the number of the minimum amount of observed variables must be equal to or greater than 05 times the number of variables in the research model (Comrey, 1973). As a result, 108 individual investors participating in direct investment in the HCMC real estate market and investors were selected according to the convenient sampling method. people who are very knowledgeable about business investment in the real estate market. Step 2: According to Nunnally and Bernstein (1994), to measure through Cronbach's confidence coefficient α must have at least three measurement variables. Cronbach coefficient α has variable value in the range [0.1]. Theoretically, the higher the Cronbach α, the better (the more reliable the scale). However, Cronbach α is too large (α> 0.95) shows that many variables in the scale are not different because they both measure a certain content in the concept, this phenomenon is called duplication in measurement. . So a scale has good reliability when it varies in the range [0.75-0.95]. If Cronbach α≥0,6 is acceptable in terms of reliability. So in this study, the author chose that scale value from 0.6 or more due to the new context and a new scale. Besides, based on the principle of avoiding duplication, variables to measure a research concept must be closely 11 correlated with each other. Therefore, when examining one correlation with the other in a concept, the item-total correlation, symbol r, is used. If a scale variable has a correlation coefficient with the total variable ≥ 0.3 then the variable is satisfactory, but if it is completely coincident (r = 1) then those two measurement variables do only one thing, and only one of two variables is sufficient. So in this study, the author chose the correlation coefficient with the total variable ≥ 0.3 to meet the requirements for the study. Step 3 (factor analysis to discover EFA): Conduct EFA analysis for all observed variables with Varimax rotation, eigenvalue> 1.0 to find out the factors representing the variables. According to Hair et al. (2010), standards when analyzing EFA include: KMO index value from 0.5 to 1, which is suitable for exploratory factor analysis, Sig <5%; The observed variables with factor loading coefficient (Factor loading) greater than 0.6 will be retained, the observed variables with factor load coefficients of less than 0.6 will be excluded from the model; Eigenvalue coefficient> 1.0. 3.1.2. Formal quantitative research Step 1 (choose sample size): For factor analysis, sample size depends on the number of observed variables included in factor analysis, according to Hair et al. (2010) the number of required observations is at least five. times the number of variables. Meanwhile, Hoang Trong and Chu Nguyen Mong Ngoc (2008), the minimum number of observed variables must be four to five times the number of variables in factor analysis. And Tabachnick and Fidell (2001) said that the minimum sample size to be achieved is calculated by the formula N ≥50 + 8m (where m is the number of independent variables). Based on the minimum number of samples of the above two groups of authors, the author chooses a sample size large enough to satisfy both above conditions with size N≥ max, corresponding to a scale of 34 observed variables, and 7 variables. independently, the minimum required number of samples is N ≥ max (50 + 8 * 7; 5 * 34) = 170 samples. To ensure the reliability of the research results, the dissertation uses 381 survey questions to put into data processing. To achieve the research objectives the author uses the method of representation and probability. Step 2 (selecting survey subjects and data collection): Sample is taken from the list of clients of real estate exchanges, real estate brokers they have successfully brokered, from the list of clients products of investors with projects, real estate located in different districts, but ensuring representative direction within the area of Ho Chi Minh City. Specifically: District 2, District 9 represents the East; Districts 6, 8, 11, Binh Tan, Tan Phu and Tan Binh represent the West; Districts 7 and 12 represent the South; District Tan Binh, Go Vap, Binh Thanh, Thu Duc represent the North. Through the list of customers who have bought the real estate, the author has contacted by phone number to ask for permission to be surveyed. The results are obtained through the following forms: direct meeting for consultation, sending questionnaire survey via email and google form. Number of questionnaires sent to people who have bought real estate is 905. The response rate is 48.8% which means that 442 responses were sent back. After screening, 61 votes were removed, of which 34 were for other purposes, 27 votes showed that respondents had never bought or sold real estate and that respondents were not fully answered. Thus, the sample of the thesis is 381 investors. Step 3 (CFA confirmation factor analysis): After the factor analysis step, the model should be tested in the next step, the CFA confirmation factor analysis step, helping to clarify the criteria scale evaluation is as follows: (i) unidimensionality; (ii) composite reliability; (iii) variance extracted; (iv) convergent validity (v) Discriminant Validity. Step 4 (analysis of SEM linear structure model): SEM structural model analysis technique: to find out the influence of psychological factors of investors on investment efficiency and impact level of factors. The method of hypothesis testing and research model using the Amos 20 tool, in addition to having advantages over traditional methods such as multivariate regression due to the limited error in regression, for SEM for the results of simultaneous relationship between variables: independent and intermediate variables; the intermediate variable and the dependent variable in the research model. Step 5 (Bootstap analysis): Bootstrap method to verify the reliability of estimates. This is a repeat sampling method and has substitution from the original N samples (Schumaker & Lomax, 1996) for B = 1,000 samples. If the mean value from the Bootstrap results with B samples and the estimated value from the N samples with the estimate of the population by the Maximum-Likelihood method (ML) tend to be close, the difference of the estimate and the difference The standard values are small and stable, allowing reliable estimates of ML from the original N samples. Step 6 (multi-group analysis): Multi-group structure analysis method to compare the research model under 12 certain groups of a qualitative variable. According to this study, the thesis seeks to compare the model showing the impact of herd psychology, overconfidence, fear of loss on investment efficiency by investment purpose groups. (to live, accumulate / do business), by loan use ratio (low / high), by investment experience (less / a lot). This method is analyzed in two models: invariant and variable. Chi-square test is used to compare between 2 models. If the Chi-square test shows that there is no difference between the invariant model and the variable model (Pvalue> 0.05), the invariant model will be selected (with higher degrees of freedom). Conversely, if the Chi_square difference is significant between the two models (Pvalue <0.05). Step 7 (testing the differences in the variables in the model): Before testing the differences of the variables, the research tests the value level of the variables compared to the population and compared to the average by One - test. Sample test, then conducted to test the differences of the variables by t-test, both tests were processed by SPSS 20 software. 3.2 Building research scales According to Creswell and Creswell (2017) in common scientific research, there are 3 ways to have scales used in research: i) Use existing scales, use primitives of scales built by previous researchers. erect; ii) Using the existing scale but supplementing and adjusted to suit the research space; iii) Build completely new scales. In this study, the author used the scale of previous studies and then discussed the experts to adjust, supplement and complete the scale in accordance with the research objectives. 3.3 Preliminary scale reliability test results by analyzing the reliability of Cronbach's Alpha Evaluate the reliability of the scale through Cronbach's Alpha coefficients with sample number 108. Of which 37 scales belong to independent variables except for the scale HIEUQUA1 and VITRI1 with the Corrected Item - Total Correlation (Corrected Item - Total Correlation) < 0.3, the author proceeds with the variable type and re-processing. The re-treatment results show that the remaining factors all have Corrected Item - Total Correlation coefficient of 0.3 and Cronbach's Alpha coefficient> 0.6, so the variables are acceptable. suitable for inclusion in the subsequent analysis. Therefore, the number of observed variables is filtered to 35 before being included in the exploratory factor analysis (EFA), as shown below: Table 3.1: Results of confidence coefficients of Cronbach's Alpha in preliminary quantitative research The concept Component notation Number of observed variables excluded Number of observed variables after processing Cronbach’s Alpha Cronbach’ s Alpha The correlation coefficient of the smallest sum Investment performance HIEU_QUA 1 3 0,838 0,709 Herd behavior BAY_DAN 0 4 0,875 0,801 Loss aversion THUA_LO 0 4 0,884 0,823 Overconfidence TU_TIN 0 4 0,890 0,819 Real estate information THONG_TI N 0 6 0,892 0,855 Real estate location VI_TRI 1 4 0,892 0,832 Macroeconomic KT_VIMO 0 4 0,757 0,593 Real estate demand CAU_BDS 0 6 0,891 0,865 Sum 2 35 Source: Author analysis and synthesis 3.4 Results of factor analysis to discover EFA in preliminary quantitative research After testing the scale by Cronbach's reliability analysis Anpha has 08 unidirectional concepts that meet the requirements to be included in the discovery factor analysis (EFA) by extracting Principal Axis Factoring and Varimax rotation. The specific analysis results are as follows: 13 Table 3.12: KMO and Barlett testing results Evaluation factors Result Condition Evaluation KMO coefficient 0,721 Từ 0,5 đến 1 Accept Sig value of the Barlett test 0,000 < 5% Accept Citation variance 73,398% > 50% Accept Eigenvalue 1,464 > 1 Accept Source: Author analysis and synthesis - KMO = 0.721, so factor analysis is appropriate; Sig (Barlett’Test) = 0,000 <5% shows that the observed variables are correlated in the overall; Eigenvalue = 1,464 represents the variation explained by each factor, the factor drawn is the best summary of information; Total variance extracted: Extraction Sums of Squared Loadings (Cumulative%) = 73,398%. This proves that the data variation of 73,398% is explained by 8 factors; Factor Loading coefficients all observed variables have values greater than 0.6, except for the observed variable KTVIMO3 whose factor load factor is less than 0.6, so this variable is not enough to explain the macroeconomic factor well. and this observed variable was excluded prior to official quantitative analysis. The results of EFA analysis show that these observed variables including 34 observed variables converge into 8 factors. Table 3.13 Results of EFA analysis Observed variables Factor Factor name TUTIN1 0,840 Overconfidence TUTIN2 0,853 TUTIN3 0,792 TUTIN4 0,743 KTVIMO1 0,867 Macroeconomic KTVIMO2 0,871 KTVIMO4 0,855 THUALO1 0,730 Loss aversion THUALO2 0,849 THUALO3 0,858 THUALO4 0,796 BAYDAN1 0,790 Herd behavior BAYDAN2 0,835 BAYDAN3 0,875 BAYDAN4 0,785 THONGTIN1 0,675 Real estate information THONGTIN2 0,.762 THONGTIN3 0,816 THONGTIN4 0,781 THONGTIN5 0,857 THONGTIN6 0,799 CAUBDS2 0,792 Real estate demand CAUBDS6 0,812 CAUBDS1 0,823 CAUBDS3 0,786 CAUBDS4 0,734 CAUBDS5 0,820 HIEUQUA2 0,801 Investment performance HIEUQUA3 0,862 HIEUQUA4 0,858 VITRI2 0,860 Real estate location VITRI3 0,867 VITRI4 0,776 VITRI5 0,895 14 CHAPTER 4: RESEARCH RESULTS AND DISCUSSION 4.1 Research results 4.1.1 Sample characteristics Primary data collected from the survey through the survey is for a period of 17 months from April 2017 to September 2018 in the Ho Chi Minh City area. The sample size is 442 individual investors, and after eliminating 61 votes, of which 34 are for other purposes, 27 votes show that respondents have never bought or sold real estate and that respondents cannot. answered in full, thus 381 votes were used for data analysis). 4.1.2 Descriptive statistics of research samples In 381 questionnaires were coded, entered and analyzed, the results of the descriptive statistics of survey participants' information were analyzed specifically as follows: Through the analysis in Table 4.1, the group of investors with little investment experience (232) showed that the number of investors with a high proportion of using loans (168) accounted for 44.1% of the total sample and investors with a high proportion of using capital. Low loans (64) accounted for 16.8% of the total sample, while the group of investors with more investment experience (149) showed that the number of investors with a low rate of using loans (109) accounted for 28.6% of the total. The number of samples and the number of investors with a high rate of using loans (40) accounts for 10.5% of the total sample, this shows that investors with little experience are often investors with high ratio of using loans and investors have more Experience is often investors have low rate of using loans. Table 4.4: Investment experience by loan use ratio group Investment experience Loan utilization rate Classification by experience Low High Less than 1 year Amount 20 75 The group of investors has little experience Accumulated 20 75 From 1 year to less than 3 years Amount 44 93 Accumulated 64 168 From 3 years to less than 5 years Amount 53 20 The group of investors has a lot of experience Accumulated 53 20 From 5 years to less than 10 years Amount 45 13 Accumulated 98 33 From 10 years or more Amount 11 7 Accumulated 109 40 Source: Author analysis and synthesis Table 4.2 shows that in the group of investors with high loan rate, the percentage of investors with business purposes (43.31%) is higher than the rate of investors with purpose to stay and accumulate (11.29%); and found that in the group of investors with a loan rate, the proportion of investors with purpose to stay and accumulate (27.82%) was higher than the rate of investors with business purposes (17.59%), meaning that investors increased Borrowing capital for business investment and investors buying real estate for living, accumulating and using capital mainly from own source. Table 4.6: Investment purpose by loan use ratio group By loan group Level Purpose Tổng Rate % Accumula ted % Business To stay, to accumulate Business To stay, to accumulate Low rate 1 13 20 33 3.41 5.25 45.41 2 26 37 63 6.82 9.71 3 28 49 77 7.35 12.86 Sum 67 106 173 17.59 27.82 High rate 4 106 29 135 27.82 7.61 54.59 5 59 14 73 15.49 3.67 Sum 165 43 208 43.31 11.29 Total 232 149 381 60.89 39.11 100 Source: Author analysis and synthesis 15 4.1.3 Results of CFA confirmation factor analysis At the CFA affirmative factor analysis step, the research conducted the evaluation steps to test the scale as follows: (i) Unidirection: Due to the suitability of the model with market data for them We have the necessary and sufficient conditions for the set of observed variables to achieve unidirection. The suitability of the model with market data is shown through the following parameters: Chi-square / df = 1,263 (0.9); TLI = 0.983 (> 0.9); CFI = 0.985 (> 0.9); RMSEA = 0.026 (<0.08); (ii) Aggregate reliability coefficient đoc measuring the concepts all ensure the condition is greater than 0.5; (iii) The variance extracted ρvc for measuring the concepts of ensuring all conditions is greater than 50%; (iv) Scale of concepts achieving convergent value when the weights of the scale are all high (> 0.5) and statistically significant (Pvalue <0.05); (v) Scale of concepts achieving discriminant value when testing whether the correlation coefficient relative to 1, with the Pvalue coefficients are both satisfied less than 5%. 4.1.4 Test research model and hypotheses 4.1.4.1 Model verification by linear model (SEM) The estimated results (standardized) of the research model show that the model is consistent with the data: CMIN / df = 1,702 ( 0.9); CFI = 0.949 (> 0.9); RMSEA = 0.034 <0.088. Estimation results of the main parameters presented show that these causal relationships are statistically significant (p <5%). Based on the above results, it is possible to conclude that the concepts in the research model are valid. 4.1.4.2 Test estimation of research model by Bootstrap (N = 1000) The author used Bootstrap method with the number of repeated samples N = 1000. The estimated results by bootstrap with N = 1000 are averaged, showing that bias appears but very small, absolute for CR <2. Therefore, it can be concluded that the estimates in the research model are reliable. 4.1.4.3 Test hypotheses with linear structure model (SEM) Estimation results of research model and bootstrap in linear structural model analysis (SEM) show that the hypothetical relationship in the official research model has statistical significance because p has high 0.049 is less than 0.05, reaching the necessary significance level (at 95% confidence level). In other words, the hypotheses in the official research model are accepted. Table 4.10: Standardized, non-standard regression coefficients of the research model Relationship Standardized Regression Weights Regression Weights Level of significance (Pvalue) TU_TIN <--- THONG_TIN 0,033 0,027 0,049 BAY_BAN <--- THONG_TIN -0,299 -0,357 0,000 THUA_LO <--- THONG_TIN -0,149 -0,194 0,007 TU_TIN <--- KT_VIMO 0,381 0,263 0,000 TU_TIN <--- VI_TRI 0,366 0,364 0,000 TU_TIN <--- CAU_BDS 0,142 0,091 0,004 HIEU_QUA <--- BAY_BAN -0,255 -0,198 0,000 HIEU_QUA <--- THUA_LO -0,368 -0,262 0,000 HIEU_QUA <--- TU_TIN 0,203 0,237 0,000 Source: Author analysis and synthesis 4.1.4.4 Multi-group analysis according to investor characteristics Based on the results of group analysis by investor characteristics (ratio of loan use, investment experience, investment purpose) shows that most of the variability analysis models are selected. From the results of the non- normalized SEM linear model, combined with the SEM result of the un-normalized deformable model for group analysis, the author synthesizes the results according to the table as follows: 16 Table 4.18: Summary of non-standardized regression results by groups of investors Relationship Regression Weights Regression coefficient according to investor characteristics Loan Experience Purpose High Low Many Few Business Stay TU_TIN <--- THONG_TIN 0,027 -0,05 0,1 0,02 0,02 0,02 0,03 BAY_BAN <--- THONG_TIN -0,357 -0,46 -0,25 -0,14 -0,45 -0,46 -0,13 THUA_LO <--- THONG_TIN -0,194 -0,14 -0,26 -0,29 -0,13 -0,12 -0,28 TU_TIN <--- KT_VIMO 0,263 0,3 0,24 0,19 0,29 0,29 0,2 TU_TIN <--- VI_TRI 0,364 0,3 0,4 0,2 0,36 0,37 0,19 TU_TIN <--- CAU_BDS 0,091 0,09 0,08 0,02 0,09 0,08 0,05 HIEU_QUA <--- BAY_BAN -0,198 -0,25 -0,04 -0,06 -0,26 -0,27 -0,05 HIEU_QUA <--- THUA_LO -0,262 -0,27 -0,1 -0,08 -0,24 -0,22 -0,1 HIEU_QUA <--- TU_TIN 0,237 0,36 -0,02 -0,33 0,26 0,29 -0,18 Source: Author analysis and synthesis 4.1.4.5 Test the differences of variables according to investor characteristics From the resulting data showing the difference in the investors' evaluation of psychological factors and behavioral and investment efficiency among groups according to the characteristics of investors presented above, the author summarizes according to the table as after: Table 4.19: Summary of results of testing the differences of psychological factors, investment efficiency Factor Test the difference b