In the third Plenary Session of the 16th CPC (Communist Party of China) Central Committee, China put forward the Scientific Outlook on Development problem, its content is comprehensive, coordinated and sustainable development. Therefore, from the perspective of sustainable development of regional economy, it is necessary to study the relationship between financial development and economic growth in a region. In recent years, the world situation has undergone profound changes, and China’s development is still in an important period of strategic opportunities, is facing a rare historical opportunity, but also facing many predictable and unpredictable risks and challenges. In order to seize the opportunity to deal with the challenge, the “12th Five-Year” plan is introduced in a timely manner. During the “12th Five-Year”, the government put the strategy on the implementation of the western development strategy in the overall strategic priority of regional development. We should give full play to the comparative advantages of all regions, promote the rational flow of regional production factors and orderly transfer of industries, and cultivate new regional economic growth poles in the Midwest. Xinjiang economic work conference has been held, nineteen provinces and cities Yuanjiang work in full swing, the preferential policies for Xinjiang have been introduced, which undoubtedly made an unprecedented foreshadowing for the regional economic development
2. Literature Review
As the core of the economy, the prospects for Xinjiang’s financial development will also be waiting for us to wait and see. The relationship between regional finance and regional economy is the extension of the relationship between finance and economy in the field of view, the research of foreign scholars usually covered by the relationship between financial development and economic growth, and on the understanding and study of experienced a long process. After the mid 17th Century, the classical school gradually flourished, they advocated economic liberalism, that a change in the quantity of money will only cause the price level in proportion to rise or fall, and not have a substantial impact on the production and supply of real output and employment, so money is neutral. Then, new classical economics believes that the currency makes the commodity exchange become more unobstructed, will not have a fundamental influence on the economy; monetary economics thinks that the money supply can affect the output and prices, but in the long term, the output is completely determined by the labor, capital, technology and other non monetary factors, money supply only determines prices. In late 20th Century, Greenwood and Jovanovic (1990) research the India capital market, the results show that only part of the funds will be converted from bank deposits into tradable securities through the stock market, stock market expansion has caused a decline in business investment in fixed assets. Stand in the perspective of rational expectations, Lucas (1988) believes that the role of Finance in promoting economic growth is limited, even can be ignored, he believes the many research “excessive emphasis” on financial factors in economic growth. Krugman (2003) who once accurate predictions about the 1998 Asian financial crisis also believes that finance is only the results is of economic growth, even it restricts economic growth, and lead to economic collapse. On the contrary, many foreign economists believe that finance can promote economic growth, and is an important factor, for example, Schumpeter (1911) , Goldsmith (1969) , Patrick (1966) etc., I have to mention that McKinnon (1973) and Shaw (1973) propose the theory of “financial deepening” and “financial inhibition”. With the development of financial mathematics and econometrics, the empirical analysis of the relationship between finance and economy by foreign economists has been more mature. For instance, Levine (1997) , Beckerman (1998) . At present, most of the foreign scholars think that the economic and financial influence each other, unfortunately, but they can not give a standard conclusion.
In our country, The research on the relationship between regional financial development and regional economic growth have not formed a theoretical system, most of them nearly indiscriminately imitate the western theory, and all of our scholars verify the existing western theory by use different methods or in different period or based on different data, innovative thinking is very lack. Of course, it also has some significance. Liu (2002) made a simple linear regression using the average cross sectional data of 31 provinces (autonomous regions and municipalities directly under the central government) in mainland China for three years in 1998-2000 years. But because of the simple linear regression, the probability of error is higher, so the result is very unreliable. Zhou and Zhong (2004) studied the relationship between financial development and economic growth based on the data of China’s central, western and eastern. Results showed that financial development and economic growth has formed a benign interaction in the eastern, while in the west, this interaction has not yet appeared. Wang (2005) analyzed the whole data of eastern and western during the period 1990-2002, his conclusion is that there exists long-term cointegration relationship between the eastern and western regions of financial development and economic growth, and financial development has obvious promotion effect on economic growth in eastern, but there is a mutual inhibition relationship between financial development and economic growth in the West. Ma (2006) use the granger-causality test respectively on the 1980-2000 data of Eastern, central and Western verification shows that significant reciprocal causation relationship exists between the eastern and central economic growth and financial development, economic growth in the western region has significantly promoted the development of finance, financial development has a certain role in promoting economic growth, but effect is not obvious, similar results are also obtained, such as Ran (2006) , Xu (2007) , etc. The more objective research comes from Wu (2009, 2010) , his analysis based on quantile regression, show that financial development on the impact of economic growth and volatility is statistically significant in different quantiles of economic growth. But there’s nothing new about it. Therefore, to sum up, we know that the study of regional economic growth and financial development by scholars in China remains at the level of verification of the existing theories in the west.
3. Empirical Analysis of Regional Financial Development and Regional Economic Growth
3.1. Object Description, Model Setting and Data Source
According to the administrative division of Xinjiang statistical yearbook, taking into account the particularity of the administrative divisions of Xinjiang, the planning of the main functional areas of Xinjiang is as follows: Urumqi, Karamay, Shihezi, Turpan area, Hami area, Changji area, Yili aera, Tacheng aera, Aletai aera area, Bazhou, Bozhou, Akesu, Kashi area, Hetian area, Altai mountain forest ecological function area, the Tarim River desert ecological function area, Altun Mountains grassland ecological function zone. Taking into account the specific situation of each functional area, the state of Yili directly under the county expanded to Yili state, so this paper chooses Urumqi, Karamay, Shihezi, Yili area, Turpan area, Hami area, Changji area, Tacheng area, Aletai area, Bozhou, Bazhou, Akesu area, Kashi area, Hetian area, total of fifteen states. We implement an empirical analysis for the relationship between financial development and economic growth in the fifteen states based on the method of quantile regression. In order to maintain the consistency of statistical indicators, the data in our study from the fifteen prefectures of Xinjiang “national economic and social development statistics bulletin 2005-2011”, and “Xinjiang Statistical Yearbook 2006-2012”, the actual time of data is 2005-2011 (detailed data are shown in the appendix). What needs to be explained is that: some of missing values and outlier, we use the method for making up missing values in time series to complement missing values and replace the outlier.
In the empirical literature of economic growth, production function is a basic estimation framework. In our empirical research, we also use it to analysis the relationship between regional financial development and economic growth, set the total production function (T) in the form of the output as a function of the abstract level of financial development and control variables, control variables are other main factors in addition to the variables thus describe the financial development level, so it can be expressed as:
Notes: represents output or added value, it is generally replaced by GDP; is the level of financial development; is the control variable.
Generally, if the elasticity research is carried out, it can be expanded on the basis of Cobb Douglas type production function. Based on the available data research the relationship between financial development and economic growth of Xinjiang’s fifteen prefectures. The explained variable take per capita GDP reflects the economic growth, with representation; explanatory variables takes two groups of variables: the financial development level and control variables. They are defined as follows:
1) The level of financial development
Financial related ratio: FIR = Financial institutions loans/GDP
Financial intermediation efficiency: FAE = Loan/Deposit
Financial savings structure: FSS = Savings/All deposits
Insurance depth: IND = Premium income/GDP
2) Control variable
Physical capital input: = Region total fixed capital/GDP
Human capital investment: = Total government expenditure/GDP
Economic openness degree: = Total export trade/GDP
Labor input: = Total wages of staff and workers/
According to the above discussion, we need to carry out the elasticity research, the model is set as follows:
3.2. Fifteen Prefectures of Xinjiang Financial Development and Economic Growth: Conditional Quantile Regression Results and Statistical Analysis
3.2.1. Comparison of Estimation Results between Conditional Median Regression and Conditional Mean Regression
2005-2011 in fifteen prefectures of Xinjiang financial development and economic growth data, including a total of 105 groups of sample data of 15 states in 7 years, relatively large sample. For comparison, conditional median regression and conditional mean regression were used for empirical analysis. This paper focuses on the similarities and differences between estimation methods, statistical tests (goodness of fit, equation significance test, significance test of variables) and the estimation of equation coefficients.
1) Estimation method
The results of conditional median regression are shown in Table 1, and the results of conditional mean regression are shown in Table 2. The estimation method of conditional quantile (median) regression and conditional mean regression is different, the conditional quantile (median) use Least absolute-deviations (LAD) estimator to estimate, the conditional mean regression use Leat squares deviations (LSD) estimator to estimate. Therefore, the estimation results are naturally different due to the different estimation methods.
2) Statistical test
The significance test (Quasi-LR test and F test) of conditional median regression and conditional mean regression were statistically significant at the significant level of 0.01. Variable significance test (t-test), in addition to the variables in Table 1, LNCKZB, LNLDZB and Table 2, the variable LNCKZB is not statistically significant at the significant level of 0.10, other variables in Table 1 and Table 2 are statistically significant at the significant level of 0.05. Because of the different calculation methods, the goodness of fit of the two estimation methods is obviously different. Generally, based on the same data, the pseudo goodness of fit (Pseudo R-squared) was significantly smaller than the goodness of fit (R-squared), and the adjusted pseudo goodness of fit (Adjusted Pseudo R-squared) was significantly smaller than the adjusted goodness of fit (Adjusted R-squared). In Table 1, Pseudo R-squared = 0.6182, Adjusted Pseudo R-squared=0.5864; In Table 2, R-squared = 0.8366, Adjusted R-squared=0.8230. In addition, in Table 2, D.W = 0.6636, because D.W. = 2 (1 − ), can calculate the = 0.6682, showed a positive correlation between the sequence order, If the AR (1) is introduced into the model setting, the goodness of fit can be improved. However, because both
Table 1. Conditional median regression results.
Note: 1) Dependent Variable: LNGDPP; 2) Method: Quantile Regression (Median); 3) Sample: 2005-2011; 4) Bootstrap method: XY-pair, reps = 200, mg = kn, seed = 673,929,944; 5) Included observations: 609.
Table 2. Conditional mean regression results.
R-squared and Adjusted R-squared have exceeded 0.80. Furthermore, there was no sequence correlation test for conditional median regression in Table 1, in order to increase the comparability, the conditional mean reversion with AR (1) is no longer given in this paper.
3) Equation coefficient estimation
The values of the conditional median regression estimator and the conditional mean regression estimator of the corresponding coefficients are obviously different. However, there is no change in the conditional median regression estimator corresponding to the eight explanatory variable coefficients and the symbol of conditional mean regression estimator. In the conditional median regression model, the estimation values of LNCKZB and LNLDZB were not significant in T-test, the rest are significant; in the conditional mean regression model, the estimation of LNCKZB is not significant, others are significant. In the conditional median regression model, the absolute value of the estimated variables LNCKZB and LNLDZB coefficients is less than the absolute value of the estimated value of the conditional mean regression explanatory variable coefficient; The absolute values of the explanatory variables LNGTZB, LNCZZB, LNFIR, LNFAE, LNFSS and LNIND coefficients in the conditional median regression model are greater than the absolute values of the estimated values of the conditional mean regression explanatory variable coefficients.
3.2.2. Analysis of the Difference and Variation of Conditional Quantile Regression Estimation Coefficients
As mentioned before, as a result of the economic gap between the various prefectures of Xinjiang, the Xinjiang economic development empirical analysis, put forward the countermeasures, it can not be generalized, not the statistical regression analysis using traditional statistical methods for the average properties. Quantile regression analysis can solve this problem. As mentioned before, because the regional economic disparity is obvious in all aspects of Xinjiang, therefore, an empirical analysis of Xinjiang’s economic development and put forward the countermeasures, we can not generalize, the traditional statistical methods can not be used for the regression analysis of the average properties. quantile regression analysis can solve this problem.
In order to reveal the impact of financial development and other control variables on economic growth at different levels of economic growth, we need to carry out conditional quantile regression estimation at different quantile levels of economic growth. Two aspects are also involved in the specific estimation: the selection of the quantile and the calculation of the standard deviation of the coefficient. First, in the selection of quantiles, 10 quantile estimates are taken here. Limited by the length of this article, only 5 quantile results are given, Namely τ = 0.10, 0.30, 0.50, 0.70, 0.90. Secondly, the standard deviation of the quantile regression coefficient is obtained by the self-help method (bootstrap) repeated sampling 200 times. The results obtained from the 5 quantiles of the 10 quantile regression are shown in Table 3.
Table 3. Quantile regression coefficient result.
1) The difference analysis of the estimators of the coefficients of different explanatory variables. In a specific quantile level of economic growth, the impact of financial development and other control variables on economic growth is different. Specifically, as shown in Table 3, the coefficients of the explanatory variables, LNGTZB, LNFAE, and LNIND, are positive in the different quantile levels of the explained variable LNGDPP, the coefficients of the explanatory variables LNCZZB, LNLDZB, LNFIR and LNFSS are negative, and the absolute value of the coefficient of the variable LNFSS is the largest at all the quantiles. The coefficient of explanatory variable LNCKZB is negative at the 0.1 and 0.9 quantiles of the explained variable. It is positive at other quantiles, and the absolute value of the negative coefficient is larger than the positive value, but its absolute value is the smallest compared with other variables. In addition, it is easy to see that the standard deviation of the quantile regression coefficient obtained by the repeated sampling of the self-help method (bootstrap) is also different from the standard deviation of the quantile regression coefficient obtained by the 200 times. For every specific quantile level of economic growth, the standard deviation of the coefficient of explanatory variable LNCKZB is the smallest, while the standard deviation of the coefficients of explanatory variables LNFAE and LNFSS is relatively large. In addition, the standard deviation of regression coefficient near the 0.5 quantile is relatively small, and the standard deviation of regression coefficient near the 0.1 and 0.9 percentile is relatively large. Notability, the tail probabilities of the significance test of the regression coefficient of the explanatory variable LNCKZB near the quantiles were more than 0.10, The tail probabilities of the significance test of the regression coefficient at the 0.1, 0.3, and 0.5 quantiles of the explanatory variable LNLDZB are all greater than 0.10. Only the P value of the tail probability of the LNFAE variable in the 0.1 and 0.9 quantile regression coefficient significant test appears more than 0.10. the top down is 0.2556 and 0.1636, respectively.
2) Variation analysis of the estimated values of the same explanatory variables. At the level of every quantile of economic growth, the impact of an explanatory variable (financial development and other control variables) on economic growth is different. As shown in Table 3, the quantile level of the dependent variable LNGDPP increases from 0.1 to 0.9. The point estimates for the coefficients of the explanatory variables (financial development and other control variables) are all changing. We focus on the analysis of the variation characteristics of the coefficient points of the variables related to the level of financial development. That is, the explanatory variables LNFIR, LNFAE, LNFSS, LNIND, specifically expressed as: The coefficient of the explanatory variable LNFIR is reduced first and then incrementing, the coefficient of LNFIR begins to increase from the 0.30 digits of LNGDPP, it basically reflects the basic law of the loan function of financial institutions with different quantile levels in LNGDPP of fifteen prefectures of Xinjiang; The coefficient of LNFAE increasing first and then decreasing, basically, the coefficient of LNFAE begins to become smaller at the 0.30 digits of the interpreted variable LNGDPP, it basically reflects the basic law of efficiency of financial intermediary role with different quantile levels in LNGDPP of fifteen prefectures of Xinjiang; The coefficient of the explanatory variable LNFSS has two sharp drops at both ends, it is gradually smaller from 0.1 to 0.3 quantiles of LNGDPP, from 0.3 to 0.7, it is progressively larger, from 0.7 to 0.9, the number is gradually reduced, This basically reflects the basic law of financial savings structure; The coefficient of the explanatory variable LNIND is reduced first and then incrementing, the coefficient of LNIND began to increase at At the 0.7 percentile of LNGDPP, which basically reflects the basic rules of the insurance industry.
4. Conclusions and Policy Recommendations
Based on the above empirical analysis, we have the following basic conclusions: there are some differences between conditional median regression and conditional mean regression. Compared with conditional mean regression, conditional (multiple) quantile regression can reveal more in-depth and comprehensive data information. Using the 2005-2011 data of fifteen prefectures of Xinjiang, the result of the conditional (multiple) quantile regression shows that the impact of financial development and other control variables on economic growth is different in specific quantile level of economic growth, showing the differences in the role of explanatory variables. In each of the different quantile levels of economic growth, the impact of an explanatory variable (financial development and other control variables) on economic growth is different, showing the volatility of the explanatory variable; Considering the level of financial development (LNFIR, LNFAE, LNFSS, LNIND) in different quantile level of economic growth, it will be found in every quantile level of economic growth, regional financial development in Xinjiang has the opposite effect on economic growth, it shows that the level of regional financial development in Xinjiang restricts the growth of the economy, unlike what we think subjectively that the promotion of Xinjiang’s financial industry, it will promote the growth of Xinjiang’s economy. It is seen from this, the predecessors think that financial development promotes economic growth, financial development will restrict economic growth or there is no relationship between financial development and economic growth is one- sided. From the data analysis results, we can see that in different stages of economic development, the level of financial development plays a different role in economic growth, and the extent of action is also different. That is, in the “budding period” of finance, economic growth needs to nurture financial development, at this time, financial development needs compensation from other sectors of the economy, at this stage, finance is a drag on economic growth; To the “growth period” of finance, financial development and economic growth are coordinated and mutually promoting, in this stage, if the two are not coordinated, there will be mutual constraints, but it is often difficult to coordinate the development of the two; If the financial level continues to develop, it will enter the “mature period”, financial development will serve the economic growth well; If finance continues to develop and enter the “period of excessive prosperity”, financial development will be a factor that restricts economic growth, and even lead to a recession and the collapse of the financial system, for example, Wall Street and Wenzhou folk lending.
4.2. Policy Recommendations
4.2.1. Efforts to Promote Leapfrog Development of Financial Scale and Efficiency with the Aid of Policy
Our state has paid more and more attention to the economic development of Xinjiang, and has launched a new round of support for Xinjiang. This provides an unprecedented opportunity for Xinjiang’s economic development. Chiefly, Xinjiang fifteen prefectures need to clear their own financial development stage, From the financial related ratio (FIR), the loans of Urumqi and Shihezi are more than its GDP, the rest are less than 1, mostly in 0.4 - 0.6, the lowest is Karamay, the financial related ratio is 0.17. Reference to Appendix Table A11. But there is no doubt that Urumqi and Karamay are the richest areas in Xinjiang, which means Urumqi relies on the total financial volume, while Karamay wins by the financial benefit (the paper defined as the gains from the financial assets of the unit). It is pointed out the development direction for other states in Xinjiang, For example, Shihezi, Hami, and Kashi with higher financial related ratios(FIR), they can continue to expand the financial scale, improve the financial related ratio and realize the leapfrog development. States with lower financial related ratios should be to expand the scale of the finance, at the same time, more efforts should be made in the financial efficiency, and the national wealth, supported by the unit loans, is constantly increasing, and the financial benefits are constantly improved, and the leap forward development is finally achieved.
4.2.2. Vigorously Advocates the Spirit of Contract, Optimizes the Environment of Indirect Financing, and Improves the Efficiency of Financial Intermediation (FAE)
Modern society is a society ruled by law, but our country is slow to establish and perfect all kinds of legal systems, in Xinjiang, where the degree of nationalization is still high, the construction of various laws and regulations is more serious and more complicated. But the modern economic society is in the era of contract economy, without the spirit of contract, it is hard for the society to have a good faith environment and lack of integrity. As a core of indirect financing business, banks will inevitably raise the loan threshold. As a result, enterprises or individuals who are short of money can hardly get the support needed, which hinders the development of enterprises and individuals, and can not provide enough power for economic growth. From the empirical results, we can see that the efficiency of financial intermediation has a significant positive impact on economic growth at each quantiles. In fact, The efficiency of financial intermediation in fifteen prefectures of Xinjiang mostly in the 30% - 55% level, no area reached the level of 70%, details see Appendix Table A12. Therefore, Xinjiang fifteen prefectures should vigorously promote the spirit of contract, to optimize the indirect financing environment, even the implementation of the national rural bank lending patterns―Five household joint insurance, and actively establish the linkage mechanism between finance, prompting banks to reduce lending threshold, in order to improve the efficiency of financial intermediation, so as to promote the regional economic growth.
4.2.3. Speed up the Establishment and Improvement of Social Security System, Reduce the Saving Levels
Macroeconomics believes that savings can provide the funds needed for economic growth, however, if the social savings rate is generally high, residents have less disposable income to consume, and lack of consumption, that is, lack of demand, which will surely lead to slow economic growth or even zero growth. It can be seen from the empirical results of this article, Xinjiang’s state financial savings structure (LNFSS) has a significant negative impact on economic growth at all the quantile levels, and the absolute value of the influence coefficient on the economy is the largest at each quantiles, it can be seen that the financial savings structure in Xinjiang has been a serious impediment to economic growth. Therefore, the reduction of the social savings rate is to provide demand for the economic development. It’s easy to think that people’s high saving concept is only to cope with unexpected needs. If the social security system is perfect, people will consume (of course, do not encourage extravagance and waste), so that economic development will be strongly supported.
4.2.4. Widely Publicized Insurance Undertakings and Accelerated the Development of Insurance
Xinjiang is a multi-ethnic region, agriculture is more developed, but agriculture has greatly affected by natural disasters, in addition, Xinjiang has three extremist forces, especially in the three southern states, three forces activity is rampant, it leads people to invest in other underdeveloped areas, but did not dare to take action. More serious is, once a major terrorist incident, the compatriots of all ethnic groups in Xinjiang mainland will quickly return home, which is very unfavorable to the economic development of Xinjiang around the state, if there is insurance, fear will be reduced or even disappear, therefore, to speed up the development of the insurance industry is very important. From the empirical results, we can see that insurance depth (LNIND) has a significant positive impact on economic growth at all quantile levels, and also has a greater impact.
Table A1. Gross domestic product (100 million yuan).
Data source: Xinjiang fifteen prefectures “national economic and social development statistics bulletin 2005-2011”.
Table A2. Per capita gross domestic product (yuan).
Data source: Xinjiang fifteen prefectures “national economic and social development statistics bulletin 2005-2011”.
Table A3. Total investment in fixed assets (billion yuan).
Data source: Xinjiang fifteen prefectures “national economic and social development statistics bulletin 2005-2011”.
Table A4. Total financial expenditure (100 million yuan).
Table A5. Total export trade (100 million yuan).
Table A6. Total wages ($100 million).
Table A7. Deposit balance (100 million yuan).
Table A8. Savings deposit (100 million yuan).
Table A9. Loan balance (billion yuan).
Table A10. Premium income (100 million yuan).
Table A11. Financial related ratio (FIR).
Note: Financial Related Ratio FIR = Financial Institutions Loan/Nominal GDP.
Table A12. Financial intermediation efficiency (FAE).
Note: Financial intermediation efficiency FAE = Loan/Deposit.
Table A13. Financial savings structure (FSS).
Note: financial savings structure FSS= resident savings/all deposits.
Table A14. The depth of insurance (IND).
Note: insurance depth IND = premium income/GDP.